Did i just made the fastest dependency checker in the world? (198 packages in 1.5 seconds)

[Mark]

Hey.

is a dependency checking time of 1.5 seconds for 189 packages fast? ^_^
and it can be ALOT faster aswell.

Page Parsed in: 1.5912721157074 seconds. (mostly php)
MySQL Parsed in: 0.0009613037109375 seconds. (MySQL only)

php is doing the recursive functions and a in_array thing (in a few arrays
with a few 100 entry`s)
so the php part can be improved for speed improvements, but the best thing
would be to write this thing in c/c++.
the mysql parsetime of all the dependency`s recursively in php is JUST
0.001(rounded up) so the total time for all this stuff could very well
be under
one second.

I did the dependency check on KDEBASE, but the test isn`t really 100% fair.
like i don`t have the currently installed packages striped out against the
ones that my dependency checker says are needed for KDEBASE. but this is
because i have very limited access to the rpm database with php. so if i
would have done this based on my current system (with the installed
packages) it would probably be a little higher parsetime for mysql because
it would need to check every package if it`s already installed on my system
and if it is it will also need to check the version and release numbers.. so
a fair estimate of the parsetime in that case would probably be 0.003 MAX
(still extremely fast)

Now why did i open up this thread? and why in the fedora-devel-list?
well i think this stuff concerns fedora and it`s (my dep checker) in
development so that`s why i posted it here.
i`m actually making this with the idea of removing yum completely (also pup,
pirut and puplet but those are still idea`s.. for that i need to learn alot
more c/c++ (no i don`t want to write those gui apps in python or even in
php)

The reason i posted it here is because i`m hoping to attract the attention
of a programmer who is willing to help me out in making this app a c/c++ app
so that it really CAN be used in fedora. or if someone could enhance the
php_rpm extension so that i can also check it against the currently
installed rpm`s instead of only playing with rpm files. or a way of reading
out the rpmdb. (i need to know the installed: Names, Version and Releases).

And why did i make it?
quite simple. i simply don`t believe that installing a package really has to
take SO MUCH time (with checking the dependency`s) so i started making my
own dependency checker. this stuff is the result of a one day php work. 2
php classes and done. the total thing is probably about 22 kilobytes!! (with
lots of useless stuff in it aswell). Originally i had the idea to make a
online dependency checker so that rpm maintainers could upload there rpm
file in the dependency checker which would check if there aren`t and
conflicting things in the header to install it.. soon after that i realized
that the original idea was fine but that so much more could be done with
this.. so adding a few nore lines changed the way the scrpt was working to
be a dependency checker only.

My initial version did the kdebase dependency`s in about 30 seconds. that
was on tables without any indexes, this one does it in 0.001 secounds and
than the tables still aren`t 100% optimal. there are still alot of doubles
in it and it isn`t splitted alphabetically (so all the A deps in a deps_a
table and so on for: 0-9 a-z.

and at this moment all the rpm files that come with fedora core 7 test 3 are
indexed.

my way of checking dependency`s has a few advantages and disadvantages.
(alot more advantages.. but i name just 3 of both)

Advantages:
- Alot faster than any rpm installation tool
- Way smaller "repodata" files (under 1MB)
- Alot less waiting on the dependency`s

Disadvantages
- A new "createrepo" program will need to be written
- Existing repository`s won`t work (perhaps will with a yet to make repo
converter which converts the sqlite or xml data to the style + format i
want)
- could devide the rpm based distributions (into distributions that use yum
and distributions that use my program)

but serious.. is this a world record? (i even think it`s faster than ubuntu
and all other systems that are currently existing)

i hope to see some reply`s on this post.
Any suggestions or critics are welcome.

here is the full output of my dependency checker:

Packages needed for: kdebase

cdparanoia-libs - required for: kdebase
glibc - required for: cdparanoia-libs
glibc-common - required for: glibc
tzdata - required for: glibc-common
-- basesystem - required for: glibc
-- setup - required for: basesystem
---- filesystem - required for: basesystem
---- libgcc - required for: glibc
-- desktop-backgrounds-basic - required for: kdebase
---- desktop-file-utils - required for: kdebase
---- glib2 - required for: desktop-file-utils
------ eject - required for: kdebase
-------- coreutils - required for: kdebase
-------- grep - required for: coreutils
-------- pcre - required for: grep
-------- libstdc++ - required for: pcre
---------- pam - required for: coreutils
---------- audit-libs - required for: pam
------------ cracklib - required for: pam
-------------- cracklib-dicts - required for: pam
---------------- libselinux - required for: pam
---------------- libsepol - required for: libselinux
------------------ mcstrans - required for: libselinux
------------------ libcap - required for: mcstrans
------------------ mktemp - required for: pam
-------------------- sed - required for: pam
---------- htdig - required for: kdebase
---------- openssl - required for: htdig
---------- e2fsprogs-libs - required for: openssl
---------- device-mapper-libs - required for: e2fsprogs-libs
------------ krb5-libs - required for: openssl
-------------- zlib - required for: openssl
------------ kdelibs - required for: kdebase
------------ arts - required for: kdelibs
------------ libX11 - required for: arts
------------ libXau - required for: libX11
-------------- libXdmcp - required for: libX11
---------------- xorg-x11-filesystem - required for: libX11
-------------- alsa-lib - required for: arts
---------------- audiofile - required for: arts
------------------ esound-libs - required for: arts
-------------------- libogg - required for: arts
---------------------- qt - required for: arts
---------------------- fontconfig - required for: qt
---------------------- freetype - required for: fontconfig
------------------------ expat - required for: fontconfig
------------------------ libICE - required for: qt
-------------------------- libSM - required for: qt
---------------------------- libXcursor - required for: qt
---------------------------- libXfixes - required for: libXcursor
------------------------------ libXrender - required for: libXcursor
------------------------------ libXext - required for: qt
-------------------------------- libXft - required for: qt
---------------------------------- libXinerama - required for: qt
------------------------------------ libXrandr - required for: qt
-------------------------------------- libjpeg - required for: qt
---------------------------------------- libmng - required for: qt
---------------------------------------- lcms - required for: libmng
---------------------------------------- libtiff - required for: lcms
------------------------------------------ libpng - required for: qt
------------------------ libvorbis - required for: arts
-------------- cups-libs - required for: kdelibs
-------------- gnutls - required for: cups-libs
-------------- libgcrypt - required for: gnutls
-------------- libgpg-error - required for: libgcrypt
---------------- hicolor-icon-theme - required for: kdelibs
------------------ xorg-x11-server-utils - required for: kdelibs
------------------ cpp - required for: xorg-x11-server-utils
-------------------- libXTrap - required for: xorg-x11-server-utils
-------------------- libXt - required for: libXTrap
---------------------- libXaw - required for: xorg-x11-server-utils
---------------------- libXmu - required for: libXaw
------------------------ libXpm - required for: libXaw
------------------------ libXfontcache - required for: xorg-x11-server-utils
-------------------------- libXi - required for: xorg-x11-server-utils
---------------------------- libXxf86misc - required for:
xorg-x11-server-utils
------------------------------ libXxf86vm - required for:
xorg-x11-server-utils
-------------------- libacl - required for: kdelibs
-------------------- libattr - required for: libacl
---------------------- libart_lgpl - required for: kdelibs
------------------------ aspell - required for: kdelibs
------------------------ aspell-en - required for: aspell
-------------------------- ncurses - required for: aspell
-------------------------- bzip2-libs - required for: kdelibs
---------------------------- libidn - required for: kdelibs
------------------------------ kdnssd-avahi - required for: kdelibs
------------------------------ avahi - required for: kdnssd-avahi
------------------------------ chkconfig - required for: avahi
-------------------------------- dbus - required for: avahi
-------------------------------- libxml2-python - required for: dbus
-------------------------------- libxml2 - required for: libxml2-python
---------------------------------- python - required for: libxml2-python
---------------------------------- db4 - required for: python
------------------------------------ gdbm - required for: python
-------------------------------------- python-libs - required for: python
---------------------------------------- readline - required for: python
------------------------------------------ sqlite - required for: python
---------------------------------- dbus-python - required for: avahi
---------------------------------- dbus-glib - required for: dbus-python
------------------------------------ initscripts - required for: avahi
------------------------------------ SysVinit - required for: initscripts
-------------------------------------- bash - required for: initscripts
---------------------------------------- udev - required for: initscripts
---------------------------------------- MAKEDEV - required for: udev
------------------------------------------ libvolume_id - required for: udev
------------------------------------------ e2fsprogs - required for:
initscripts
------------------------------------------ device-mapper - required for:
e2fsprogs
-------------------------------------------- ethtool - required for:
initscripts
---------------------------------------------- mingetty - required for:
initscripts
------------------------------------------------ module-init-tools -
required for: initscripts
-------------------------------------------------- util-linux - required
for: initscripts
-------------------------------------------------- ConsoleKit-libs -
required for: util-linux
---------------------------------------------------- popt - required for:
util-linux
---------------------------------------------------- net-tools - required
for: initscripts
------------------------------------------------------ syslog-ng - required
for: initscripts
------------------------------------------------------ eventlog - required
for: syslog-ng
-------------------------------------------------------- tcp_wrappers-libs -
required for: syslog-ng
---------------------------------------------------------- logrotate -
required for: syslog-ng
-------------------------------------------------------- sysklogd - required
for: initscripts
-------------------------------------- libdaemon - required for: avahi
-------------------------------- avahi-qt3 - required for: kdnssd-avahi
-------------------------------- libutempter - required for: kdelibs
-------------------------------- shadow-utils - required for: libutempter
---------------------------------- libxslt - required for: kdelibs
------------------------------------ perl - required for: kdelibs
------------------------------------ mod_perl-devel - required for: perl
------------------------------------ httpd-devel - required for:
mod_perl-devel
------------------------------------ apr-devel - required for: httpd-devel
------------------------------------ apr - required for: apr-devel
-------------------------------------- pkgconfig - required for: apr-devel
-------------------------------------- apr-util-devel - required for:
httpd-devel
-------------------------------------- apr-util - required for:
apr-util-devel
-------------------------------------- openldap - required for: apr-util
-------------------------------------- cyrus-sasl-lib - required for:
openldap
---------------------------------------- db4-devel - required for:
apr-util-devel
------------------------------------------ expat-devel - required for:
apr-util-devel
-------------------------------------------- openldap-devel - required for:
apr-util-devel
-------------------------------------------- cyrus-sasl-devel - required
for: openldap-devel
---------------------------------------- httpd - required for: httpd-devel
---------------------------------------- gawk - required for: httpd
-------------------------------------- mod_perl - required for:
mod_perl-devel
-------------------------------------- perl-BSD-Resource - required for:
mod_perl
---------------------------------------- perl-devel - required for: mod_perl
-------------------------------------- redhat-artwork - required for:
kdelibs
-------------------------------------- gtk2-engines - required for:
redhat-artwork
-------------------------------------- gtk2 - required for: gtk2-engines
-------------------------------------- atk - required for: gtk2
---------------------------------------- cairo - required for: gtk2
------------------------------------------ pango - required for: gtk2
------------------------------------------ libthai - required for: pango
---------------------------------------- fedora-logos - required for:
redhat-artwork
---------------------------------------- redhat-menus - required for:
kdelibs
-------------- Glide3-libGL - required for: kdebase
-------------- mesa-libGL - required for: Glide3-libGL
-------------- libdrm - required for: mesa-libGL
---------------- Glide3 - required for: Glide3-libGL
---------------- policycoreutils - required for: Glide3
---------------- audit-libs-python - required for: policycoreutils
---------------- kernel-headers - required for: audit-libs-python
------------------ libselinux-python - required for: policycoreutils
-------------------- libsemanage - required for: policycoreutils
---------------- mesa-libGLU - required for: kdebase
------------------ libXcomposite - required for: kdebase
-------------------- libXdamage - required for: kdebase
---------------------- libXScrnSaver - required for: kdebase
------------------------ libXtst - required for: kdebase
-------------------------- libfontenc - required for: kdebase
---------------------------- hal - required for: kdebase
---------------------------- ConsoleKit - required for: hal
------------------------------ acl - required for: hal
-------------------------------- cryptsetup-luks - required for: hal
---------------------------------- dmidecode - required for: hal
------------------------------------ libusb - required for: hal
-------------------------------------- pm-utils - required for: hal
-------------------------------------- kbd - required for: pm-utils
-------------------------------------- usermode - required for: kbd
-------------------------------------- libuser - required for: usermode
---------------------------------------- passwd - required for: usermode
---------------------------------------- pciutils - required for: pm-utils
---------------------------------------- hwdata - required for: pciutils
------------------------------ libraw1394 - required for: kdebase
-------------------------------- lm_sensors - required for: kdebase
-------------------------------- libsysfs - required for: lm_sensors
---------------------------------- libsmbclient - required for: kdebase
------------------------------------ libxkbfile - required for: kdebase
-------------------------------------- psmisc - required for: kdebase

Total RPMS to install: 186

Dependency Conflicts: 0



Page Parsed in: 1.5912721157074 seconds.

MySQL Parsed in: 0.0009613037109375 seconds.

[seth vidal]

On Mon, 2007-04-09 at 22:46 +0200, Mark wrote:
>  Hey.
>  
>  is a dependency checking time of 1.5 seconds for 189 packages fast?
>  ^_^
>  and it can be ALOT faster aswell.
>  
>  Page Parsed in: 1.5912721157074 seconds. (mostly php)
>  MySQL Parsed in: 0.0009613037109375 seconds. (MySQL only) 
>   
Unless I misread this you're not performing any files-as-dep look ups.
So any time a package requires /foo/bar/baz you have to look that up in
the file listing of each package to see what provides it. this
information isn't listed in the rpm -q --provides output

-sv

[Mark]

i don`t really understand what you mean..
i actually check only all the requires of a package to the provides of the
packages that are in the database.
but the files-as-dep.. i don`t understand that.

2007/4/9, seth vidal <skvidal(a)linux.duke.edu&gt;:
>  On Mon, 2007-04-09 at 22:46 +0200, Mark wrote:
>  > Hey.
>  >
>  > is a dependency checking time of 1.5 seconds for 189 packages fast?
>  > ^_^
>  > and it can be ALOT faster aswell.
>  >
>  > Page Parsed in: 1.5912721157074 seconds. (mostly php)
>  > MySQL Parsed in: 0.0009613037109375 seconds. (MySQL only)
>  >
>
>  Unless I misread this you're not performing any files-as-dep look ups.
>  So any time a package requires /foo/bar/baz you have to look that up in
>  the file listing of each package to see what provides it. this
>  information isn't listed in the rpm -q --provides output
>
>  -sv
>
>
>  --
>  fedora-devel-list mailing list
>  fedora-devel-list(a)redhat.com
>  https://www.redhat.com/mailman/listinfo/fedora-devel-list
>

[seth vidal]

On Mon, 2007-04-09 at 23:19 +0200, Mark wrote:
>  i don`t really understand what you mean..
>  i actually check only all the requires of a package to the provides of
>  the packages that are in the database.
>  but the files-as-dep.. i don`t understand that.
>   
Does your provides database include every file provided by every
package?

so that if you encounter:

Requires: /some/path/to/a/file

it can resolve it?

also how are you doing your versioned-dep calculation? Are you calling
out to rpm for the rpm version comparisons?


so if a pkg requires:
 foo > 1:1.2

how are you checking to see if the package foo-1.3-1 provides that
dependency or not?

-sv

[Mark]

>  Does your provides database include every file provided by every
>  package?
>
>  so that if you encounter:
>
>  Requires: /some/path/to/a/file
>
>  it can resolve it? 

if there is another package that has "/some/path/to/a/file" in it`s Provides
i can resolve it. otherwise i can`t. (yet)

also how are you doing your versioned-dep calculation? Are you calling
>  out to rpm for the rpm version comparisons? 

that`s not that hard at all.
i just have a table with the package id, package provides/requires and the
version of the requires/provides though this can use some tweaking.

how are you checking to see if the package foo-1.3-1 provides that
>  dependency or not? 

to put it simple:
SELECT * FROM PROVIDES WHERE provide_id = '##' AND provide_name = 'foo-1.3-1
';
(simple sample.. it`s not the query that i actually use.. that`s alot longer
:P)



2007/4/9, seth vidal <skvidal(a)linux.duke.edu&gt;:
>
> On Mon, 2007-04-09 at 23:19 +0200, Mark wrote:
> > i don`t really understand what you mean..
> > i actually check only all the requires of a package to the provides of
> > the packages that are in the database.
> > but the files-as-dep.. i don`t understand that.
> >
>  Does your provides database include every file provided by every
>  package?
>
>  so that if you encounter:
>
>  Requires: /some/path/to/a/file
>
>  it can resolve it? 
>
> also how are you doing your versioned-dep calculation? Are you calling
>  out to rpm for the rpm version comparisons? 
>
>
> so if a pkg requires:
> foo > 1:1.2
>
> how are you checking to see if the package foo-1.3-1 provides that
>  dependency or not? 
>
> -sv
>
>
>
> --
> fedora-devel-list mailing list
> fedora-devel-list(a)redhat.com
> https://www.redhat.com/mailman/listinfo/fedora-devel-list
>

[Matthew Miller]

On Mon, Apr 09, 2007 at 11:34:23PM +0200, Mark wrote:
>  also how are you doing your versioned-dep calculation? Are you
> calling
>  >out to rpm for the rpm version comparisons?
>  that`s not that hard at all.
>  i just have a table with the package id, package provides/requires and the
>  version of the requires/provides though this can use some tweaking. 
What Seth is getting at is that this is harder than you think. Once you
start doing it in the complete way (which is required for correct answers),
you'll loose speed.


>  how are you checking to see if the package foo-1.3-1 provides that
>  >dependency or not?
>  to put it simple:
>  SELECT * FROM PROVIDES WHERE provide_id = '##' AND provide_name = 'foo-1.3-1
How do you know if version 1.3 is greater or less than epoch 1 version 1.2?

-- 
Matthew Miller           mattdm(a)mattdm.org          <http://mattdm.org/>
Boston University Linux      ------>              <http://linux.bu.edu/>

[Mark]

>  That won't work.
>
>  That only checks for the precise provide, not for the ranges.
>
>  example:
>
>  bar requires foo > 1.1
>
>  you only have foo 1.3
>
>  now, implicitly in the above is that foo 1.3 fulfills the dep of foo >
>  1.1
>
>  So you have to check to see that the ranges overlap.
>
>  Do you understand now? 

i get it now :) i didn`t even thought about that one.

What Seth is getting at is that this is harder than you think. Once you
>  start doing it in the complete way (which is required for correct
>  answers),
>  you'll loose speed. 

is there anything more to check besides the requires and provides with the
versions?
because if the version thing is all that needs to be fixed (besides cross
checking the already installed packages) than the speed will get lower but
still well below a second (for mysql) i think 0.002 :P and this is with
tables that have over 40.000 entry's!! (the first time i even work with
tables that big

hmm.. it will get complex if you also want things like:
Foo is installed, but has version 1.1
Bar requires Foo version 1.3. Do you want me to update Foo?

and for those paths.. is it possible that no package provides the path but
that it`s already installed (so simply not in the provides list) if that`s
the case the final c/c++ script will have to work together with slocate..
and when a slocate database isn`t made yet (with updatedb) it will take alot
of the speed...or not... first load some stuff in the memory, than run
updatedb and while that`s running do the calculations on the things that are
stored in the memory.. the idea`s keep comming :P


And to make sure i have the right idea..
imagine i want to instakk "FooBar" than the actions i need to do are:

1. first check what`s needed (with versions)
2. cross reference (1) against the currently installed packages (with
versions)
3. Check for avalible updates on (1) (versions only)
4. install all: new packages, updates and install for dependency

right?


2007/4/9, Matthew Miller <mattdm(a)mattdm.org&gt;:
>
> On Mon, Apr 09, 2007 at 11:34:23PM +0200, Mark wrote:
> > also how are you doing your versioned-dep calculation? Are you calling
> > >out to rpm for the rpm version comparisons?
> > that`s not that hard at all.
> > i just have a table with the package id, package provides/requires and
> the
> > version of the requires/provides though this can use some tweaking.
>
> What Seth is getting at is that this is harder than you think. Once you
>  start doing it in the complete way (which is required for correct
>  answers),
>  you'll loose speed. 
>
>
> > how are you checking to see if the package foo-1.3-1 provides that
> > >dependency or not?
> > to put it simple:
> > SELECT * FROM PROVIDES WHERE provide_id = '##' AND provide_name = '
> foo-1.3-1
>
> How do you know if version 1.3 is greater or less than epoch 1 version 1.2
> ?
>
> --
> Matthew Miller           mattdm(a)mattdm.org          <http://mattdm.org/>
> Boston University Linux      ------>              <http://linux.bu.edu/>
>
> --
> fedora-devel-list mailing list
> fedora-devel-list(a)redhat.com
> https://www.redhat.com/mailman/listinfo/fedora-devel-list
>

[Matthew Miller]

On Tue, Apr 10, 2007 at 12:13:31AM +0200, Mark wrote:
>  > What Seth is getting at is that this is harder than you think.
> Once you
>  >start doing it in the complete way (which is required for correct
>  >answers), you'll loose speed.
>  is there anything more to check besides the requires and provides with the
>  versions? 
All sorts of stuff. Check the yum source code. For starters the other things
mentioned already:, epochs, and the fact that all packages are considered to
be providing all files they contain, even with dependencies. But consider
this too: RPM's notion of version comparison is a bit unintuitive when it
comes to mixing letters and numbers -- is 1.2b3 greater or less than 1.20?
What about 1.2.0? And for that matter, is 1.9 higher than 1.10?


-- 
Matthew Miller           mattdm(a)mattdm.org          <http://mattdm.org/>
Boston University Linux      ------>              <http://linux.bu.edu/>

[Mark]

it`s only logic to assume that 1.10 is higher than 1.9...  but for the
1.2b3things.. those will require another check.

2007/4/10, Matthew Miller <mattdm(a)mattdm.org&gt;:
>  On Tue, Apr 10, 2007 at 12:13:31AM +0200, Mark wrote:
>  > > What Seth is getting at is that this is harder than you think. Once
>  you
>  > >start doing it in the complete way (which is required for correct
>  > >answers), you'll loose speed.
>  > is there anything more to check besides the requires and provides with
>  the
>  > versions?
>
>  All sorts of stuff. Check the yum source code. For starters the other
>  things
>  mentioned already:, epochs, and the fact that all packages are considered
>  to
>  be providing all files they contain, even with dependencies. But consider
>  this too: RPM's notion of version comparison is a bit unintuitive when it
>  comes to mixing letters and numbers -- is 1.2b3 greater or less than 1.20?
>  What about 1.2.0? And for that matter, is 1.9 higher than 1.10?
>
>
>  --
>  Matthew Miller           mattdm(a)mattdm.org          <http://mattdm.org/>
>  Boston University Linux      ------>              <http://linux.bu.edu/>
>
>  --
>  fedora-devel-list mailing list
>  fedora-devel-list(a)redhat.com
>  https://www.redhat.com/mailman/listinfo/fedora-devel-list
>

[Ralf Ertzinger]

Hi.

On Tue, 10 Apr 2007 08:37:26 +0200, Mark wrote:

>  it`s only logic to assume that 1.10 is higher than 1.9...  but for
> the
>  1.2b3things.. those will require another check. 
This is the RPM depsolver. Logic will not get you anywhere :)

[Matthew Miller]

On Tue, Apr 10, 2007 at 08:37:26AM +0200, Mark wrote:
>  it`s only logic to assume that 1.10 is higher than 1.9...  but for
> the 
Logic, maybe -- a numeric comparison, no.


-- 
Matthew Miller           mattdm(a)mattdm.org          <http://mattdm.org/>
Boston University Linux      ------>              <http://linux.bu.edu/>

[Mark]

if it isn`t logic than why is the current system still existing?
it can be so much faster is logic is being used ^_^


2007/4/10, Matthew Miller <mattdm(a)mattdm.org&gt;:
>  On Tue, Apr 10, 2007 at 08:37:26AM +0200, Mark wrote:
>  > it`s only logic to assume that 1.10 is higher than 1.9...  but for the
>
>  Logic, maybe -- a numeric comparison, no.
>
>
>  --
>  Matthew Miller           mattdm(a)mattdm.org          <http://mattdm.org/>
>  Boston University Linux      ------>              <http://linux.bu.edu/>
>
>  --
>  fedora-devel-list mailing list
>  fedora-devel-list(a)redhat.com
>  https://www.redhat.com/mailman/listinfo/fedora-devel-list
>

[Nicolas Mailhot]

Le Mar 10 avril 2007 16:23, Mark a écrit :
>  if it isn`t logic than why is the current system still existing?
>  it can be so much faster is logic is being used ^_^ 
RPM versionning rules have to match versionning rules used by the packaged
projects in real life. Users want to match package versions to the
versions advertised on upstream web sites.

Since what software writers consider sane versionning is only slightly
more predictable than winning loto numbers, rpm internal rules border on
the complex.

In other news the software industry was still US-centric, and americans
still pondering the benefits of going metric.

-- 
Nicolas Mailhot

[Mark]

it wouldn`t be a problem if you could say things like this:

1.2b3 > 1.2b4 and that just the changing number is compared to see if a
higher one is avalible.
so in this case that woud be this:

3 > 4

perhaps that`s possible with regular expressions..??? but than again you
can`t check in mysql to see if the version (with a "b" in it) is greather or
less than a existing version.

i will do my best to get this working ^_^ i would love to have the fastest
dependency checker in the world that`s realy working.. even without the
strange logic of version numbers.

to be clear:
1.01
1.10
1.1
== smaler than:
1.20

that`s how my dep checker will work.
and anyone that would be willing write this script in c++ with me?



2007/4/10, Nicolas Mailhot <nicolas.mailhot(a)laposte.net&gt;:
>  Le Mar 10 avril 2007 16:23, Mark a écrit :
>  > if it isn`t logic than why is the current system still existing?
>  > it can be so much faster is logic is being used ^_^
>
>  RPM versionning rules have to match versionning rules used by the packaged
>  projects in real life. Users want to match package versions to the
>  versions advertised on upstream web sites.
>
>  Since what software writers consider sane versionning is only slightly
>  more predictable than winning loto numbers, rpm internal rules border on
>  the complex.
>
>  In other news the software industry was still US-centric, and americans
>  still pondering the benefits of going metric.
>
>  --
>  Nicolas Mailhot
>
>  --
>  fedora-devel-list mailing list
>  fedora-devel-list(a)redhat.com
>  https://www.redhat.com/mailman/listinfo/fedora-devel-list
>

[Matthew Miller]

On Tue, Apr 10, 2007 at 04:54:08PM +0200, Mark wrote:
>  to be clear:
>  1.01
>  1.10
>  1.1
>  == smaler than:
>  1.20
>  that`s how my dep checker will work. 
Which is smaller: 5.6 or 5.00504?

-- 
Matthew Miller           mattdm(a)mattdm.org          <http://mattdm.org/>
Boston University Linux      ------>              <http://linux.bu.edu/>

[Brian Wheeler]

On Tue, 2007-04-10 at 11:19 -0400, Matthew Miller wrote:
>  On Tue, Apr 10, 2007 at 04:54:08PM +0200, Mark wrote:
>  > to be clear:
>  > 1.01
>  > 1.10
>  > 1.1
>  > == smaler than:
>  > 1.20
>  > that`s how my dep checker will work.
>  
>  Which is smaller: 5.6 or 5.00504?
>   
Perl, right?  5.00504, because it came out before 5.6. However, it seems
to be broken in terms of version number rules -- which is why they
switched to 5.6.  5.00504 was essentially 5.5 patch 4.

You may look around the perl archives for perl's version number
comparison stuff to see how they handled it.


Brian

>  -- 
>  Matthew Miller           mattdm(a)mattdm.org          <http://mattdm.org/>
>  Boston University Linux      ------>              <http://linux.bu.edu/>
>

[Matthew Miller]

On Tue, Apr 10, 2007 at 04:54:08PM +0200, Mark wrote:
>  to be clear:
>  1.01
>  1.10
>  1.1
>  == smaler than:
>  1.20
>  that`s how my dep checker will work. 
Also, I need to point out that if your dep checker works like the above,
it's not right. First, it's got 1.1 and 1.10 in the wrong order. And, 1.01
and 1.1 are, in fact, *equal*. Yeah, good times with RPM.


-- 
Matthew Miller           mattdm(a)mattdm.org          <http://mattdm.org/>
Boston University Linux      ------>              <http://linux.bu.edu/>

[Mark]

well.. that list is just written out of nowhere :P i don`t actually see how
the dep checker is doing it. (i can echo it if i want)

Which is smaller: 5.6 or 5.00504?


i would say that 5.00504 = smaller than 5.6 though i`m not sure if mysql
thinks the same..

And, 1.01
>  and 1.1 are, in fact, *equal*. 

normally it isn`t.. in this sample.. it is :s that`s because you also have a
1.10...
oh my god.. how much speed am i gonna lose in this :P



2007/4/10, Matthew Miller <mattdm(a)mattdm.org&gt;:
>  On Tue, Apr 10, 2007 at 04:54:08PM +0200, Mark wrote:
>  > to be clear:
>  > 1.01
>  > 1.10
>  > 1.1
>  > == smaler than:
>  > 1.20
>  > that`s how my dep checker will work.
>
>  Also, I need to point out that if your dep checker works like the above,
>  it's not right. First, it's got 1.1 and 1.10 in the wrong order. And, 1.01
>  and 1.1 are, in fact, *equal*. Yeah, good times with RPM.
>
>
>  --
>  Matthew Miller           mattdm(a)mattdm.org          <http://mattdm.org/>
>  Boston University Linux      ------>              <http://linux.bu.edu/>
>
>  --
>  fedora-devel-list mailing list
>  fedora-devel-list(a)redhat.com
>  https://www.redhat.com/mailman/listinfo/fedora-devel-list
>

[Matthew Miller]

On Tue, Apr 10, 2007 at 08:49:14PM +0200, Mark wrote:
>  Which is smaller: 5.6 or 5.00504?
>  i would say that 5.00504 = smaller than 5.6 though i`m not sure if mysql
>  thinks the same.. 
But for a depsolver, all that matters is what RPM thinks.


>  And, 1.01
>  >and 1.1 are, in fact, *equal*.
>  normally it isn`t.. in this sample.. it is :s that`s because you also have a
>  1.10... 
What does the 1.10 have to do with it? One is still one, which isn't ten.




-- 
Matthew Miller           mattdm(a)mattdm.org          <http://mattdm.org/>
Boston University Linux      ------>              <http://linux.bu.edu/>

[Jesse Keating]

On Tuesday 10 April 2007 14:49:14 Mark wrote:
>  Which is smaller: 5.6 or 5.00504?
>
>
>  i would say that 5.00504 = smaller than 5.6 though i`m not sure if mysql
>  thinks the same.. 
Actually, 5.6 is smaller.  5 and 5 is compared, then 00504 and 6 is compared.  
00504 actually becomes 504 which is higher than just 6 and thus wins.

-- 
Jesse Keating
Release Engineer: Fedora

[Mark]

so.. is mysql yust removing out those first 2 zero`s?... i`m having a hard
time believing that..
oke.. i wrote a tiny php script to make this clear:

<?php

    if (5.6 < 5.00504)
    {
        echo 'not what i would have expected<br />';
    }
    else
    {
        echo 'see, 5.00504 is smaller than 5.6 like i said<br />';
    }

    if (1.1 == 1.01)
    {
        echo 'not what i would have expected<br />';
    }
    else
    {
        echo 'see, they 1.1 is not equal to 1.01<br />';
    }

?>

the result on my php (5.2.1) is: "see, 5.00504 is smaller than 5.6 like i
said" and "see, they 1.1 is not equal to 1.01"
those are the facts that are gonna be used in the dep checker. though i will
probably need to break the recursive loop quite a few time if those strange
version numbers like 1.1b2 occur...
so jesse... how is your math going? ;)

2007/4/10, Jesse Keating <jkeating(a)redhat.com&gt;:
>  On Tuesday 10 April 2007 14:49:14 Mark wrote:
>  > Which is smaller: 5.6 or 5.00504?
>  >
>  >
>  > i would say that 5.00504 = smaller than 5.6 though i`m not sure if mysql
>  > thinks the same..
>
>  Actually, 5.6 is smaller.  5 and 5 is compared, then 00504 and 6 is
>  compared.
>  00504 actually becomes 504 which is higher than just 6 and thus wins.
>
>  --
>  Jesse Keating
>  Release Engineer: Fedora
>
>  --
>  fedora-devel-list mailing list
>  fedora-devel-list(a)redhat.com
>  https://www.redhat.com/mailman/listinfo/fedora-devel-list
>
>

[Jesse Keating]

On Tuesday 10 April 2007 16:01:53 Mark wrote:
>  the result on my php (5.2.1) is: "see, 5.00504 is smaller than
> 5.6 like i
>  said" and "see, they 1.1 is not equal to 1.01"
>  those are the facts that are gonna be used in the dep checker. though i
>  will probably need to break the recursive loop quite a few time if those
>  strange version numbers like 1.1b2 occur...
>  so jesse... how is your math going? ;) 
Perhaps you're not getting it.  RPM does not compare things such as
"5.00504" 
vs "5.6", rather it breaks each part separated by a period into an individual 
item and compares those items.  In effect, it compares 5 to 5, then 00504 to 
6.  Given that 00504 (which is really 504) is a larger number than 6 (or 
00006 if you rather), the 5.00504 package is rpm newer than the 5.6 package.

-- 
Jesse Keating
Release Engineer: Fedora

[Andy Burns]

On 10/04/07, Mark <markg85(a)gmail.com&gt; wrote:

>  so.. is mysql yust removing out those first 2 zero`s?... i`m having a
> hard time believing that..
>  oke.. i wrote a tiny php script to make this clear: 
>  the result on my php (5.2.1) is: "see, 5.00504 is smaller than
> 5.6 like i said" and "see, 
they  1.1 is not equal to 1.01"

It isn't what PHP *thinks*, it it what RPM *knows* that counts.

>  those are the facts that are gonna be used in the dep checker. though
> i will probably need to break the recursive loop quite a few time if those strange version
> numbers like 1.1b2 occur...
>  so jesse... how is your math going? ;) 
Errr, people have been very patient and provided gentle hints about
how many things you've overlooked, you need to start reading the
source code of yum and rpm I think.

[drago01]

On 4/10/07, Mark <markg85(a)gmail.com&gt; wrote:
>  so.. is mysql yust removing out those first 2 zero`s?... i`m having a hard
>  time believing that..
>  oke.. i wrote a tiny php script to make this clear:
>
>  <?php
>
>      if (5.6 < 5.00504)
>      {
>          echo 'not what i would have expected<br />';
>      }
>      else
>      {
>          echo 'see, 5.00504 is smaller than 5.6 like i said<br />';
>      }
>
>      if (1.1 == 1.01)
>      {
>          echo 'not what i would have expected<br />';
>      }
>      else
>      {
>          echo 'see, they 1.1 is not equal to 1.01<br />';
>      }
>
>  ?> 


it was not about 1.1 and 1.01 but 1.10 should be bigger than 1.1 (which is
the same if you do == )
this can be solved if you split the version string and compare each part
(explode() )


the result on my php (5.2.1) is: "see, 5.00504 is smaller than 5.6 like i
>  said" and "see, they 1.1 is not equal to 1.01"
>  those are the facts that are gonna be used in the dep checker. though i
>  will probably need to break the recursive loop quite a few time if those
>  strange version numbers like 1.1b2 occur...
>  so jesse... how is your math going? ;)
>
>  2007/4/10, Jesse Keating <jkeating(a)redhat.com&gt;:
>  >
>  > On Tuesday 10 April 2007 14:49:14 Mark wrote:
>  > > Which is smaller: 5.6 or 5.00504?
>  > >
>  > >
>  > > i would say that 5.00504 = smaller than 5.6 though i`m not sure if
>  > mysql
>  > > thinks the same..
>  >
>  > Actually, 5.6 is smaller.  5 and 5 is compared, then 00504 and 6 is
>  > compared.
>  > 00504 actually becomes 504 which is higher than just 6 and thus wins.
>  >
>  > --
>  > Jesse Keating
>  > Release Engineer: Fedora
>  >
>  > --
>  > fedora-devel-list mailing list
>  > fedora-devel-list(a)redhat.com
>  > https://www.redhat.com/mailman/listinfo/fedora-devel-list
>  >
>  >
>
>  --
>  fedora-devel-list mailing list
>  fedora-devel-list(a)redhat.com
>  https://www.redhat.com/mailman/listinfo/fedora-devel-list
>

[Nicolas Mailhot]

Le mardi 10 avril 2007 à 22:01 +0200, Mark a écrit :
>  so.. is mysql yust removing out those first 2 zero`s?... i`m having
> a
>  hard time believing that..
>  oke.. i wrote a tiny php script to make this clear: 
I also suggest

<?php
    if (2.6.21 > 2.6.3)
    {
        echo 'not what i would have expected<br />'; 
    }
    else
    {
        echo 'I'll fix your kernel now!<br />';
    }


-- 
Nicolas Mailhot

[Toshio Kuratomi]

On Tue, 2007-04-10 at 22:01 +0200, Mark wrote:
>  so.. is mysql yust removing out those first 2 zero`s?... i`m having
> a
>  hard time believing that..
>  oke.. i wrote a tiny php script to make this clear:
>  
>  <?php
>  
>      if (5.6 < 5.00504)
>      {
>          echo 'not what i would have expected<br />'; 
>      }
>      else
>      {
>          echo 'see, 5.00504 is smaller than 5.6 like i said<br />';
>      }
>  
>      if (1.1 == 1.01)
>      {
>          echo 'not what i would have expected<br />'; 
>      }
>      else
>      {
>          echo 'see, they 1.1 is not equal to 1.01<br />';
>      }
>  
>  ?>
>  
>  the result on my php (5.2.1) is: "see, 5.00504 is smaller than 5.6
>  like i said" and "see, they 1.1 is not equal to 1.01"
>  those are the facts that are gonna be used in the dep checker. though
>  i will probably need to break the recursive loop quite a few time if
>  those strange version numbers like 1.1b2 occur... 
>  so jesse... how is your math going? ;)
>   
Mark, what people are trying to point out to you is that it doesn't
matter what php, mysql, you, jesse, or I think is a more recent version.
All that matters is what rpm thinks is a more recent version since rpm
is what will ultimately be managing the packages on the system.

Your two tests illustrate this problem.  The tests show that php has one
idea of what version is larger which will disagree with what rpm finds.
Therefore it will be returning wrong values.

-Toshio

PS: Your tests aren't accurately testing what php will need to check
either.  Your tests are using php floating point types but versions
(even without throwing rpm into the mix) aren't simple numbers.  2.10.1
vs 2.10.2 for instance.  And when you can compare those properly, try
comparing 2.10.1 vs 2.10.02.  And when you can compare those try
comparing 2.10.1beta1 vs 2.10.1... ad nauseum.

[Matthew Miller]

On Tue, Apr 10, 2007 at 10:01:53PM +0200, Mark wrote:
>  the result on my php (5.2.1) is: "see, 5.00504 is smaller than
> 5.6 like i
>  said" and "see, they 1.1 is not equal to 1.01"
>  those are the facts that are gonna be used in the dep checker. though i will 
Let me step back a bit. What are you trying to accomplish with this dep
checker?



-- 
Matthew Miller           mattdm(a)mattdm.org          <http://mattdm.org/>
Boston University Linux      ------>              <http://linux.bu.edu/>

[Mark]

check dependency`s :)
ultimately.. a new yum version which just takes a second or 2 to check the
dependency`s.

Perhaps you're not getting it.  RPM does not compare things such as "5.00504
>  "
>  vs "5.6", rather it breaks each part separated by a period into an
>  individual
>  item and compares those items.  In effect, it compares 5 to 5, then 00504
>  to
>  6.  Given that 00504 (which is really 504) is a larger number than 6 (or
>  00006 if you rather), the 5.00504 package is rpm newer than the 5.6package. 

first off.. sorry for not understanding the rpm part.. i understand it now.
rpm clearly breaks all logic in version numbering. (atleast my point of view
for the version logic)

a question: what i still don`t get is where rpm is gonna step is.. aren`t i
just checking the dependency`s with php + mysql (c++ + mysql when it`s all
finished) and just use rpm to install / remove stuff?
that was what i had in mind.

if that isn`t possible than perhaps a patch should be written for rpm so
that it doesn`t check for versions anymore.. it just installs nomather what
you put in (version wise) and let my dep checker do all the version stuff.

also i didn`t have the idea to let rpm be the package manager.. just to
install rpm`s, remove rpm`s and get info about rpm`s. my repo checker will
do all the math work.

to put it simple.. rpm has to do this:
- install rpms
- remove rpms
- don`t ask questions and just DO what i tell it to do
- maintain the rpmdb
- get info about a rpm (if needed)

The dep checker needs to do:
- read the rpmdb to see what rpm`s are installed
- generate a list of new rpm`s that have to be installed (lets say for when
you want to install kdebase)
- send the install commands to rpm (which in turn adds the info to the
database and installs the rpm)

done?
this should be about it right?



2007/4/10, Matthew Miller <mattdm(a)mattdm.org&gt;:
>  On Tue, Apr 10, 2007 at 10:01:53PM +0200, Mark wrote:
>  > the result on my php (5.2.1) is: "see, 5.00504 is smaller than 5.6 like
>  i
>  > said" and "see, they 1.1 is not equal to 1.01"
>  > those are the facts that are gonna be used in the dep checker. though i
>  will
>
>  Let me step back a bit. What are you trying to accomplish with this dep
>  checker?
>
>
>
>  --
>  Matthew Miller           mattdm(a)mattdm.org          <http://mattdm.org/>
>  Boston University Linux      ------>              <http://linux.bu.edu/>
>
>  --
>  fedora-devel-list mailing list
>  fedora-devel-list(a)redhat.com
>  https://www.redhat.com/mailman/listinfo/fedora-devel-list
>

[Richard Hughes]

On Tue, 2007-04-10 at 22:52 +0200, Mark wrote:
>  if that isn`t possible than perhaps a patch should be written for
> rpm
>  so that it doesn`t check for versions anymore.. it just installs
>  nomather what you put in (version wise) and let my dep checker do all
>  the version stuff.  
Not to be rude, but could you start a sourceforce project and then get a
mailing list to discuss this sort of stuff in private?

This isn't really a issue for fedora core development list, well, until
you have a drop in replacement for yum (which might be a little while
off I think).

Thanks.

Richard.

[Mark]

>  You could do that. But, do you really want your program coming up with
>  different answers than the standard tools?
>  
if the "standard tools" are using logic that`s just no logic than yes
someone needs to make a change

This isn't really a issue for fedora core development list, well, until
>  you have a drop in replacement for yum (which might be a little
> while
>  off I think). 

you can say that (last part) again... it could take a long while till it`s
done.. considering that i just know what pointers are in c++ and besides the
pointers i know a few other basics but nothing interesting enough to be able
to make this stuff in c++ and that`s the exact reason why i`m first making
it in php.

Not to be rude, but could you start a sourceforce project and then get a
>  mailing list to discuss this sort of stuff in private? 

not for now.

This isn't really a issue for fedora core development list, well, until
>  you have a drop in replacement for yum (which might be a little
> while
>  off I think). 

Well i need to visions of the people here to make this stuff work so that`s
one reason why i posted it here.
i could make a sourceforge account and create a mailing list there but it`s
highly unlikely that i would have gotten as much feedback as i get here. and
in the end it`s all for fedora do why not post it here. (perhaps the
fedora-list would have been a better place)



2007/4/10, Richard Hughes <hughsient(a)gmail.com&gt;:
>
> On Tue, 2007-04-10 at 22:52 +0200, Mark wrote:
> > if that isn`t possible than perhaps a patch should be written for rpm
> > so that it doesn`t check for versions anymore.. it just installs
> > nomather what you put in (version wise) and let my dep checker do all
> > the version stuff.
>
> Not to be rude, but could you start a sourceforce project and then get a
>  mailing list to discuss this sort of stuff in private? 
>
> This isn't really a issue for fedora core development list, well, until
>  you have a drop in replacement for yum (which might be a little
> while
>  off I think). 
>
> Thanks.
>
> Richard.
>
>
> --
> fedora-devel-list mailing list
> fedora-devel-list(a)redhat.com
> https://www.redhat.com/mailman/listinfo/fedora-devel-list
>

[Richard Hughes]

On Tue, 2007-04-10 at 23:38 +0200, Mark wrote:
>          You could do that. But, do you really want your program
> coming
>          up with 
>          different answers than the standard tools?
>   
>  if the "standard tools" are using logic that`s just no logic than yes
>  someone needs to make a change 
No, new tools have to be backwards compatible else it breaks the entire
toolchain that's been working for years.

>          This isn't really a issue for fedora core development
> list,
>          well, until
>          you have a drop in replacement for yum (which might be a
>          little while 
>          off I think).
>  
>  you can say that (last part) again... it could take a long while till
>  it`s done.. considering that i just know what pointers are in c++ and
>  besides the pointers i know a few other basics but nothing interesting
>  enough to be able to make this stuff in c++ and that`s the exact
>  reason why i`m first making it in php.  
So if you admit you have little programming knowledge and not even a
working prototype, can you PLEASE stop posting to this list. 

>          Not to be rude, but could you start a sourceforce project
> and
>          then get a
>          mailing list to discuss this sort of stuff in private?
>  
>  not for now. 
Then you are going to make yourself very unpopular.

>          This isn't really a issue for fedora core development
> list,
>          well, until
>          you have a drop in replacement for yum (which might be a
>          little while
>          off I think).
>  
>  Well i need to visions of the people here to make this stuff work so
>  that`s one reason why i posted it here. 
>  i could make a sourceforge account and create a mailing list there but
>  it`s highly unlikely that i would have gotten as much feedback as i
>  get here. and in the end it`s all for fedora do why not post it here.
>  (perhaps the fedora-list would have been a better place)  
To be honest, I think you need to read some source code of rpm and yum
(and a good book on program design) before you start posting claims
about a super fast depsolver.

Sorry if this sounds harsh.

Richard.

[Mark]

lol

>  No, new tools have to be backwards compatible else it breaks the
> entire
>  toolchain that's been working for years. 

i will find a way to "fix" that issue.. perhaps using yum if the
"good" way
can`t resolve the dependency`s

So if you admit you have little programming knowledge and not even a
>  working prototype, can you PLEASE stop posting to this list. 

do i need to have c++ knowledge to make a prototype? yum isn`t written in
c++ (or it wasn`t.. i dont really know there current coding state). i have
alot of php knowledge and that will do fine for the moment but i`m also
trying to learn c++ ^_^

Then you are going to make yourself very unpopular.


that`s mean... i need to start somewhere. and without feedback i`m not
getting far.

To be honest, I think you need to read some source code of rpm and yum
>  (and a good book on program design) before you start posting claims
>  about a super fast depsolver. 

i didn`t exactly claim that i have the fastest dependency checker.. it was
written as a question. but judging from all the feedback i got here i need
to improve alot to make the timings more accurate compared to yum. but it`s
still likely to be under 1 second for the mysql parsing. and i can look
through the yum and rpm sources but that won`t do much good because i won`t
understand 99% of it.. the other 1% is the include, print and if statement
stuff ;)

Sorry if this sounds harsh.


kinda.. you could just give me suggestions and help me out with the code.






2007/4/10, Richard Hughes <hughsient(a)gmail.com&gt;:
>
> On Tue, 2007-04-10 at 23:38 +0200, Mark wrote:
> >         You could do that. But, do you really want your program coming
> >         up with
> >         different answers than the standard tools?
> >
> > if the "standard tools" are using logic that`s just no logic than yes
> > someone needs to make a change
>
>  No, new tools have to be backwards compatible else it breaks the
> entire
>  toolchain that's been working for years. 
>
> >         This isn't really a issue for fedora core development list,
> >         well, until
> >         you have a drop in replacement for yum (which might be a
> >         little while
> >         off I think).
> >
> > you can say that (last part) again... it could take a long while till
> > it`s done.. considering that i just know what pointers are in c++ and
> > besides the pointers i know a few other basics but nothing interesting
> > enough to be able to make this stuff in c++ and that`s the exact
> > reason why i`m first making it in php.
>
> So if you admit you have little programming knowledge and not even a
>  working prototype, can you PLEASE stop posting to this list.
>
> >         Not to be rude, but could you start a sourceforce project and
> >         then get a
> >         mailing list to discuss this sort of stuff in private?
> >
> > not for now.
>
> Then you are going to make yourself very unpopular.
>
> >         This isn't really a issue for fedora core development list,
> >         well, until
> >         you have a drop in replacement for yum (which might be a
> >         little while
> >         off I think).
> >
> > Well i need to visions of the people here to make this stuff work so
> > that`s one reason why i posted it here.
> > i could make a sourceforge account and create a mailing list there but
> > it`s highly unlikely that i would have gotten as much feedback as i
> > get here. and in the end it`s all for fedora do why not post it here.
> > (perhaps the fedora-list would have been a better place)
>
> To be honest, I think you need to read some source code of rpm and yum
>  (and a good book on program design) before you start posting claims
>  about a super fast depsolver. 
>
> Sorry if this sounds harsh.
>
> Richard.
>
>
> --
> fedora-devel-list mailing list
> fedora-devel-list(a)redhat.com
> https://www.redhat.com/mailman/listinfo/fedora-devel-list
>

[Matthew Miller]

On Tue, Apr 10, 2007 at 11:38:39PM +0200, Mark wrote:
>  >You could do that. But, do you really want your program coming up
> with
>  >different answers than the standard tools?
>  if the "standard tools" are using logic that`s just no logic than yes
>  someone needs to make a change 
Well, the difficult part here is: that logic is what is used in all of the
packages for which you're trying to calculate. Therefore, using any other
scheme, no matter how much better in some objective sense, just plain won't
work. It might work for a *new* distribution designed around your new logic,
but that's no help for Fedora, 

Additionally, as someone else noted, the versioning schemes used in real
packages out in the world are often crazy too -- and in some cases, 1.11 is
greater than 1.2 (because 1.1, 1.2, 1.3 ... 1.9, 1.10, 1.11) and in others
it's less (because 1.0, 1.01, 1.02 ... 1.1, 1.11, 1.12 ... 1.2). No matter
how clever you make the code, there's no good way to deal with that -- so
might as well pick some way and be consistent. Which is what we've got.


>  This isn't really a issue for fedora core development list, well,
> until
>  >you have a drop in replacement for yum (which might be a little while
>  >off I think). 
I don't think the above is quite fair -- developing tools for Fedora is
certainly on topic. Even for non-advanced programmers.

-- 
Matthew Miller           mattdm(a)mattdm.org          <http://mattdm.org/>
Boston University Linux      ------>              <http://linux.bu.edu/>

[seth vidal]

On Tue, 2007-04-10 at 19:55 -0400, Matthew Miller wrote:

>  > This isn't really a issue for fedora core development list,
> well, until
>  > >you have a drop in replacement for yum (which might be a little while
>  > >off I think).
>  
>  I don't think the above is quite fair -- developing tools for Fedora is
>  certainly on topic. Even for non-advanced programmers.
>   
+1.

Alternative depsolvers and discussion on how we should be doing that
should definitely happen here or on other package-mgmt-related lists.
However, a fair bit of this discussion would be best suited by various
parties reading about rpm's  version comparison and yum's dependency
resolution methods before commenting on their virtue or vice.

-sv

[Mark]

would it work if i predefine all versions in another table (or column) in
the way that rpm does it? that way i don`t lose any speed and i`m still
compatible with yum ^_^ sounds like a wondefull solution to me.

only problem left is how to deal with things like: 1.1b2 or 1.1stable
or do i need to convert the b and stable things to numbers but in a unique
way so that it doesn`t effect the rpm version stuff?

sample:

a = 001100
b = 002200
c = 003300

stable = 011110
unstable = 022220

or i can imagine that rpm gets stucg again if it founds a version like this:
1.b2
cause if i`m right rpm will see that (if i do my cenvertion) as:
1.002200 = 1.22 ?

would this work?



2007/4/11, seth vidal <skvidal(a)linux.duke.edu&gt;:
>  On Tue, 2007-04-10 at 19:55 -0400, Matthew Miller wrote:
>
>  > > This isn't really a issue for fedora core development list, well,
>  until
>  > > >you have a drop in replacement for yum (which might be a little while
>  > > >off I think).
>  >
>  > I don't think the above is quite fair -- developing tools for Fedora is
>  > certainly on topic. Even for non-advanced programmers.
>  >
>
>  +1.
>
>  Alternative depsolvers and discussion on how we should be doing that
>  should definitely happen here or on other package-mgmt-related lists.
>  However, a fair bit of this discussion would be best suited by various
>  parties reading about rpm's  version comparison and yum's dependency
>  resolution methods before commenting on their virtue or vice.
>
>  -sv
>
>
>
>  --
>  fedora-devel-list mailing list
>  fedora-devel-list(a)redhat.com
>  https://www.redhat.com/mailman/listinfo/fedora-devel-list
>

[Olivier Galibert]

On Wed, Apr 11, 2007 at 09:01:33AM +0200, Mark wrote:
>  would this work? 
You just have to do what rpm does, whatever your implementation
language is.  I join what rpm does.  I think that code is under the
GPL (it's directly part of the rpm source in any case, check there if
you want to see the license).

Incidentally, the version comparison is not the worse part by far.
It's the file dependencies that are the real bitch.

  OG.

[Valent Turkovic]

>  I don't think the above is quite fair -- developing tools for
> Fedora is
>  certainly on topic. Even for non-advanced programmers.
>  
I agree!

-- 
http://kernelreloaded.blog385.com/
linux, blog, anime, spirituality, windsurf, wireless
registered as user #367004 with the Linux Counter, http://counter.li.org.
ICQ: 2125241
Skype: valent.turkovic

[Matthias Saou]

Mark wrote :

>  you can say that (last part) again... it could take a long while till
> it`s
>  done.. considering that i just know what pointers are in c++ and besides the
>  pointers i know a few other basics but nothing interesting enough to be able
>  to make this stuff in c++ and that`s the exact reason why i`m first making
>  it in php. 
Then you definitely want to start by coding a PHP module in C which
would wrap around rpmlib. You could then map rpm files to objects in PHP
to get all the info you need from the rpm files (requires and provides
as well as file lists for instance) and obviously create useful
functions, of which the first one could be to compare versions... and
would be 100% accurate since it would be using rpm's internal version
comparison.

PHP5 is now also quite good at parsing XML, so playing with repodata
shouldn't be hard either.

Good luck! :-)

Matthias

-- 
Clean custom Red Hat Linux rpm packages : http://freshrpms.net/
Fedora Core release 6 (Zod) - Linux kernel 2.6.20-1.2943.fc6
Load : 0.59 0.57 0.55

[Mark]

thanx for that.
at the moment i`ve used this to do the rpm stuff:
http://cekirdek.uludag.org.tr/~meren/php_rpm/
i`m using that to get all the stuff out of the rpm files.


>  PHP5 is now also quite good at parsing XML, so playing with repodata
>  shouldn't be hard either. 

i`m not sure if it would be wise (for me) to try and improve the existing
repodata structure.. also it`s just to big for what it provides.. and i want
to do all the stuff my way first before i improve other things.

And i think i found a way to bypass rpm`s off way of checking for version
numbers.

imagine you have:
a = 1001001
b = 1002001

1.1b2 = 1.1[1002001]2 == right for rpm`s version check? (i think it is)
1.b2 = 1.[1002001] == right for rpm`s version check? (i think it is)

in this case 1.1b2 ranks higher than 1.b2(rare that it even exists but
well.. you gotta count it in)

Now what am i doing?
quite simple. version stuff can`t handle anything different than numbers so
i need to convert
alphabetic things to numbers and it MUST start other than 0 and and other
than 0 otherwise
RPM will is gonna play tricks with me.

if this is possible it can be even as fast as the current measured speed (
0.001 second)
and why? well.. i just need to recalculate all versions in a php/c/c++ and
put it in another
table or column (column would be best in this case). than i can just change
the query slightly
and it`s checking on rpm compatible version numbers ^_^

now for another rpm version number question.

imagine the following 2 versions:
0.2.2
1.0.3

with rpm will the versions get (because it`s striping zerro`s):
2.2
1.3
??

or or is it just striping out zerro`s when there is more than 1 zerro?
i think it won`t change the initial versions in this case.

or when the length is moer than 1 character (i didn`t had a
change to look over the rpmlib coding)

some more info on the versions like how they are calculated
and in which case would be helpfull.. or a link to the file
that i need to look through




2007/4/11, Matthias Saou <
thias(a)spam.spam.spam.spam.spam.spam.spam.egg.and.spam.freshrpms.net&gt;:
>
> Mark wrote :
>
> > you can say that (last part) again... it could take a long while till
> it`s
> > done.. considering that i just know what pointers are in c++ and besides
> the
> > pointers i know a few other basics but nothing interesting enough to be
> able
> > to make this stuff in c++ and that`s the exact reason why i`m first
> making
> > it in php.
>
> Then you definitely want to start by coding a PHP module in C which
> would wrap around rpmlib. You could then map rpm files to objects in PHP
> to get all the info you need from the rpm files (requires and provides
> as well as file lists for instance) and obviously create useful
> functions, of which the first one could be to compare versions... and
> would be 100% accurate since it would be using rpm's internal version
> comparison.
>
>  PHP5 is now also quite good at parsing XML, so playing with repodata
>  shouldn't be hard either. 
>
> Good luck! :-)
>
> Matthias
>
> --
> Clean custom Red Hat Linux rpm packages : http://freshrpms.net/
> Fedora Core release 6 (Zod) - Linux kernel 2.6.20-1.2943.fc6
> Load : 0.59 0.57 0.55
>
> --
> fedora-devel-list mailing list
> fedora-devel-list(a)redhat.com
> https://www.redhat.com/mailman/listinfo/fedora-devel-list
>

[Matthias Saou]

Mark wrote :

>  some more info on the versions like how they are calculated
>  and in which case would be helpfull.. or a link to the file
>  that i need to look through 
Someone already answered that one, I'm afraid : The rpm source code.
You can speculate all you want on "replacement" methods to compare
versions, but chances you'll get it right are near zero... unless you
use the exact same algorithm (and trick and hacks?) than rpm itself.

Matthias

-- 
Clean custom Red Hat Linux rpm packages : http://freshrpms.net/
Fedora Core release 6 (Zod) - Linux kernel 2.6.20-1.2943.fc6
Load : 2.17 2.18 2.37

[Mark]

guess where this "hacking" reminds me of..
the hacking i need to do with css stylesheets to make them Internet Explorer
compatible!
and now this needs to be done in linux aswell.. even for one of it`s core
components.

o well.. i`m gonna browse through that source as soon as i get home.


2007/4/11, Matthias Saou <
thias(a)spam.spam.spam.spam.spam.spam.spam.egg.and.spam.freshrpms.net&gt;:
>  Mark wrote :
>
>  > some more info on the versions like how they are calculated
>  > and in which case would be helpfull.. or a link to the file
>  > that i need to look through
>
>  Someone already answered that one, I'm afraid : The rpm source code.
>  You can speculate all you want on "replacement" methods to compare
>  versions, but chances you'll get it right are near zero... unless you
>  use the exact same algorithm (and trick and hacks?) than rpm itself.
>
>  Matthias
>
>  --
>  Clean custom Red Hat Linux rpm packages : http://freshrpms.net/
>  Fedora Core release 6 (Zod) - Linux kernel 2.6.20-1.2943.fc6
>  Load : 2.17 2.18 2.37
>
>  --
>  fedora-devel-list mailing list
>  fedora-devel-list(a)redhat.com
>  https://www.redhat.com/mailman/listinfo/fedora-devel-list
>

[Ralf Ertzinger]

Hi.

On Wed, 11 Apr 2007 12:50:17 +0200, Mark wrote:

>  and now this needs to be done in linux aswell.. even for one of it`s
>  core components. 
RPM has been called many things, but never pretty.

[Matthew Miller]

On Tue, Apr 10, 2007 at 10:52:39PM +0200, Mark wrote:
>  a question: what i still don`t get is where rpm is gonna step is..
> aren`t i
>  just checking the dependency`s with php + mysql (c++ + mysql when it`s all
>  finished) and just use rpm to install / remove stuff?
>  that was what i had in mind.
>  if that isn`t possible than perhaps a patch should be written for rpm so
>  that it doesn`t check for versions anymore.. it just installs nomather what
>  you put in (version wise) and let my dep checker do all the version stuff. 
You could do that. But, do you really want your program coming up with
different answers than the standard tools?

Particularly, packagers work to make sure that the dependency chains work
with RPM's logic (strange as it is in some cases). If your system doesn't
work the same way, all bets are off.



-- 
Matthew Miller           mattdm(a)mattdm.org          <http://mattdm.org/>
Boston University Linux      ------>              <http://linux.bu.edu/>

[seth vidal]

On Mon, 2007-04-09 at 23:34 +0200, Mark wrote:
>          Does your provides database include every file provided by
>          every
>          package?
>          
>          so that if you encounter: 
>          
>          Requires: /some/path/to/a/file
>          
>          it can resolve it?
>  
>  if there is another package that has "/some/path/to/a/file" in it`s
>  Provides i can resolve it. otherwise i can`t. (yet)
>  
>          also how are you doing your versioned-dep calculation? Are you
>          calling
>          out to rpm for the rpm version comparisons?
>  
>  that`s not that hard at all.
>  i just have a table with the package id, package provides/requires and
>  the version of the requires/provides though this can use some
>  tweaking.
>  
>          how are you checking to see if the package foo-1.3-1 provides
>          that
>          dependency or not?
>  
>  to put it simple:
>  SELECT * FROM PROVIDES WHERE provide_id = '##' AND provide_name =
>  'foo-1.3-1'; 
>  (simple sample.. it`s not the query that i actually use.. that`s alot
>  longer :P)  

That won't work.

That only checks for the precise provide, not for the ranges.

example:

bar requires foo > 1.1

you only have foo 1.3

now, implicitly in the above is that foo 1.3 fulfills the dep of foo >
1.1

So you have to check to see that the ranges overlap.

Do you understand now?

-sv