Hello guys,
I took over the maintenance of urw-fonts to update them to the latest version (ghostscript and other packages needs them). However, there are several problems in versioning of it, and I would like to discuss future steps for the package.
The current problems: * our urw-fonts are kind of old (~2 years), and ghostscript is unable to build without latest version correctly (unless I patch it, and it would still cause ghostscript crashes for some instances AFAIK)
* according to git log the urw-fonts in Fedora should be based on version 1.10, but the source code archive indicates version 1.07 pre-release (1.06 in fact, if we check the source code)
* the NVR of urw-fonts does not correlate to this at all - urw-fonts-2.4-22.fc24.noarch
* we obtain the source archive from Artifex (aka upstream responsible for ghostscript), nowadays they are no longer using the X.Y.Z versioning system. Instead, they have created a separate git repository for it, and are doing "snapshot based" releases... Their latest release is: urw-base35-20160926.zip
* I asked them if they could go back to X.Y.Z versioning system, here's what Chris Liddell replied:
"If you check the versions in the font files, you'll see why I switched to the release dates. For several releases they never changed from 1.10, and with the latest release, they're back to 1.00. Since this is how URW++ release them to us, there's not a huge amount we can do."
* and as you can see above, they also changed the name from 'urw-fonts' to 'urw-base35' because of this
As you can see, the versioning of urw-fonts is total mess right now, and I would like to bring back some order to it, but I don't want it to backfire on me because of how URW++ tends to version their fonts... Here's my proposed solution to this: - create a new package for urw-fonts which would obsolete the current urw-fonts - choose a similar name to the new package - 'urw-base35-fonts' or similar
And either... 1) stick to URW++ based versioning - this would mean (at the moment) Version == 1.0 and adding snapshot == 20160926 (YYYYMMDD as described in https://fedoraproject.org/wiki/Packaging:Versioning#Snapshot_packages) 2) or map the X.Y.Z versioning to YYYYMMDD from upstream - IOW - X == Year, Y == Month, Z == Day (based on the snapshot date in the name of the source archive) 3) or set the Version to some constant (35 for example) and just use the snapshot to distinguish between older and newer releases.
I am affraid that I would pick option 1) it could pose problems in the future again, because there's not guarantee that URW++ will follow sane versioning. So, personally, I would choose 2) or 3).
There's also one more option, and that is to base the package on upstream's git repository and the snapshot scheme, because we would be using snapshot string in the package name anyway. And it would also solve one more issue that upstream is not shipping license files in the archive. (I have already contacted to correct this.)
What are you thoughts, guys? Anyone has a better idea how to solve this mess? Or which option would you recommend?
Thank you in advance for all your ideas.
Best regards,
David Kaspar [Dee'Kej] *Associate Software Engineer* *Brno, Czech Republic*
RED HAT | TRIED. TESTED. TRUSTED. Every airline in the Fortune 500 relies on Red Hat. Find out why at Trusted | Red Hat http://www.redhat.com/en/about/trusted.
On Wednesday, 30 November 2016 at 12:14, David Kaspar [Dee'Kej] wrote:
I took over the maintenance of urw-fonts to update them to the latest version (ghostscript and other packages needs them). However, there are several problems in versioning of it, and I would like to discuss future steps for the package.
The current problems:
[...]
- we obtain the source archive from Artifex (aka upstream responsible for
ghostscript), nowadays they are no longer using the X.Y.Z versioning system. Instead, they have created a separate git repository for it, and are doing "snapshot based" releases... Their latest release is: urw-base35-20160926.zip
Do you mean this location? http://downloads.ghostscript.com/public/fonts/
- I asked them if they could go back to X.Y.Z versioning system, here's
what Chris Liddell replied:
"If you check the versions in the font files, you'll see why I switched to the release dates. For several releases they never changed from 1.10, and with the latest release, they're back to 1.00. Since this is how URW++ release them to us, there's not a huge amount we can do."
That makes sense.
- and as you can see above, they also changed the name from 'urw-fonts' to
'urw-base35' because of this
I also found this: http://git.ghostscript.com/?p=urw-core35-fonts.git;a=summary and they're called urw-core35-fonts here. It might be wise to ask upstream to clarify before choosing the new package name.
As you can see, the versioning of urw-fonts is total mess right now, and I would like to bring back some order to it, but I don't want it to backfire on me because of how URW++ tends to version their fonts... Here's my proposed solution to this:
- create a new package for urw-fonts which would obsolete the current
urw-fonts
- choose a similar name to the new package - 'urw-base35-fonts' or similar
Good idea, but see above.
And either...
- stick to URW++ based versioning - this would mean (at the moment)
Version == 1.0 and adding snapshot == 20160926 (YYYYMMDD as described in https://fedoraproject.org/wiki/Packaging:Versioning#Snapshot_packages)
I'm not sure where your Version == 1.0 comes from. If they're versioned only by date now, then you have two options. Use Version: 0 in the new package in anticipation of upstream eventually reintroducing semantic versioning. Or, Version: YYYYMMDD. Admittedly, the latter looks nicer and upstream already said they'll stick to it.
- or map the X.Y.Z versioning to YYYYMMDD from upstream - IOW - X == Year,
Y == Month, Z == Day (based on the snapshot date in the name of the source archive) 3) or set the Version to some constant (35 for example) and just use the snapshot to distinguish between older and newer releases.
What does the number "35" mean here, anyway?
I am affraid that I would pick option 1) it could pose problems in the future again, because there's not guarantee that URW++ will follow sane versioning. So, personally, I would choose 2) or 3).
If you worry about upstream versioning sanity, then stick with Version: 0 and follow the snapshot versioning guidelines.
There's also one more option, and that is to base the package on upstream's git repository and the snapshot scheme, because we would be using snapshot string in the package name anyway. And it would also solve one more issue that upstream is not shipping license files in the archive. (I have already contacted to correct this.)
The exact location of the source doesn't matter too much as long as it's official and pristine. I agree it might be better to use the git repo directly since it contains both the licence indication and its full text.
I hope the above helps. You might want to review the current font packaging guidelines as well: https://fedoraproject.org/wiki/Packaging:FontsPolicy .
Regards, Dominik
On 30/11/2016, 9:38 PM, "Dominik 'Rathann' Mierzejewski" dominik@greysector.net wrote:
> 2) or map the X.Y.Z versioning to YYYYMMDD from upstream - IOW - X == Year, > Y == Month, Z == Day (based on the snapshot date in the name of the source > archive) > 3) or set the Version to some constant (35 for example) and just use the > snapshot to distinguish between older and newer releases.
What does the number "35" mean here, anyway?
Each PostScript specification defines a core set of fonts: - PostScript 1 has 13 base fonts. - PostScript 2 has 35 base fonts. - PostScript 3 has 135 base fonts.
So, the URW++ base35 font pack is a replacement for the 35 Adobe PostScript Level 2 base fonts.
Doug
On Wednesday, 30 November 2016 at 14:17, Douglas Kosovic wrote:
On 30/11/2016, 9:38 PM, "Dominik 'Rathann' Mierzejewski" dominik@greysector.net wrote:
- or map the X.Y.Z versioning to YYYYMMDD from upstream - IOW - X == Year,
Y == Month, Z == Day (based on the snapshot date in the name of the source archive) 3) or set the Version to some constant (35 for example) and just use the snapshot to distinguish between older and newer releases.
What does the number "35" mean here, anyway?
Each PostScript specification defines a core set of fonts:
- PostScript 1 has 13 base fonts.
- PostScript 2 has 35 base fonts.
- PostScript 3 has 135 base fonts.
So, the URW++ base35 font pack is a replacement for the 35 Adobe PostScript Level 2 base fonts.
Thanks for the explanation. It would be useful to put this information in the Summary of the new package if not in the name. Definitely not in the version field as it has nothing to do with versioning.
Regards, Dominik
On Wed, Nov 30, 2016 at 12:38 PM, Dominik 'Rathann' Mierzejewski < dominik@greysector.net> wrote:
I also found this: http://git.ghostscript.com/?p=urw-core35-fonts.git;a=summary and they're called urw-core35-fonts here. It might be wise to ask upstream to clarify before choosing the new package name.
OK, will do.
I'm not sure where your Version == 1.0 comes from. If they're versioned
only by date now, then you have two options. Use Version: 0 in the new package in anticipation of upstream eventually reintroducing semantic versioning. Or, Version: YYYYMMDD. Admittedly, the latter looks nicer and upstream already said they'll stick to it.
The Version == 1.0 comes from the source code of the fonts themselves. Running 'grep "Version" *.afm' tells me that there are all files with Version == 1.0, except two of them (which have Version
If you worry about upstream versioning sanity, then stick with Version: 0 and follow the snapshot versioning guidelines.
There's also one more option, and that is to base the package on
upstream's
git repository and the snapshot scheme, because we would be using
snapshot
string in the package name anyway. And it would also solve one more issue that upstream is not shipping license files in the archive. (I have
already
contacted to correct this.)
The exact location of the source doesn't matter too much as long as it's official and pristine. I agree it might be better to use the git repo directly since it contains both the licence indication and its full text.
Upstream has heard to my request and fixed it. ( http://bugs.ghostscript.com/show_bug.cgi?id=697390)%E2%80%8B
And yes, what Douhlas wrote is correct (about the 35 fonts), and I will have that noted in the %description section.
Anyway, since determining the Version field is still unclear, I think the most sense to me right now is to proceed with option 2) - IOW - to bypass the versioning from URW++ completely, and have Version field based on snapshot string, in a way: X.Y.Z == YYYY.MM.DD
Or do you some problem with this approach?
Thanks! :) David Kaspar [Dee'Kej] *Associate Software Engineer* *Brno, Czech Republic*
RED HAT | TRIED. TESTED. TRUSTED. Every airline in the Fortune 500 relies on Red Hat. Find out why at Trusted | Red Hat http://www.redhat.com/en/about/trusted.
On Wednesday, 30 November 2016 at 16:07, David Kaspar [Dee'Kej] wrote:
On Wed, Nov 30, 2016 at 12:38 PM, Dominik 'Rathann' Mierzejewski < dominik@greysector.net> wrote:
[...]
I'm not sure where your Version == 1.0 comes from. If they're versioned
only by date now, then you have two options. Use Version: 0 in the new package in anticipation of upstream eventually reintroducing semantic versioning. Or, Version: YYYYMMDD. Admittedly, the latter looks nicer and upstream already said they'll stick to it.
The Version == 1.0 comes from the source code of the fonts themselves. Running 'grep "Version" *.afm' tells me that there are all files with Version == 1.0, except two of them (which have Version
If they don't have the same version then it doesn't make sense to use the version of *some* of them as base.
If you worry about upstream versioning sanity, then stick with Version: 0 and follow the snapshot versioning guidelines.
There's also one more option, and that is to base the package on upstream's git repository and the snapshot scheme, because we would be using snapshot string in the package name anyway. And it would also solve one more issue that upstream is not shipping license files in the archive. (I have already contacted to correct this.)
The exact location of the source doesn't matter too much as long as it's official and pristine. I agree it might be better to use the git repo directly since it contains both the licence indication and its full text.
Upstream has heard to my request and fixed it. ( http://bugs.ghostscript.com/show_bug.cgi?id=697390)
And yes, what Douhlas wrote is correct (about the 35 fonts), and I will have that noted in the %description section.
Anyway, since determining the Version field is still unclear, I think the most sense to me right now is to proceed with option 2) - IOW - to bypass the versioning from URW++ completely, and have Version field based on snapshot string, in a way: X.Y.Z == YYYY.MM.DD
Or do you some problem with this approach?
As I said, please do not invent the version on your own. Please apply the existing snapshot guidelines instead, i.e.: Version: 0
Release: 0.N.YYYYMMDD or Release: 0.N.YYYYMMDDgitHASH
Regards, Dominik
On Wed, Nov 30, 2016 at 05:35:19PM +0100, Dominik 'Rathann' Mierzejewski wrote:
On Wednesday, 30 November 2016 at 16:07, David Kaspar [Dee'Kej] wrote:
On Wed, Nov 30, 2016 at 12:38 PM, Dominik 'Rathann' Mierzejewski < dominik@greysector.net> wrote:
[...]
I'm not sure where your Version == 1.0 comes from. If they're versioned
only by date now, then you have two options. Use Version: 0 in the new package in anticipation of upstream eventually reintroducing semantic versioning. Or, Version: YYYYMMDD. Admittedly, the latter looks nicer and upstream already said they'll stick to it.
The Version == 1.0 comes from the source code of the fonts themselves. Running 'grep "Version" *.afm' tells me that there are all files with Version == 1.0, except two of them (which have Version
If they don't have the same version then it doesn't make sense to use the version of *some* of them as base.
If you worry about upstream versioning sanity, then stick with Version: 0 and follow the snapshot versioning guidelines.
There's also one more option, and that is to base the package on upstream's git repository and the snapshot scheme, because we would be using snapshot string in the package name anyway. And it would also solve one more issue that upstream is not shipping license files in the archive. (I have already contacted to correct this.)
The exact location of the source doesn't matter too much as long as it's official and pristine. I agree it might be better to use the git repo directly since it contains both the licence indication and its full text.
Upstream has heard to my request and fixed it. ( http://bugs.ghostscript.com/show_bug.cgi?id=697390)
And yes, what Douhlas wrote is correct (about the 35 fonts), and I will have that noted in the %description section.
Anyway, since determining the Version field is still unclear, I think the most sense to me right now is to proceed with option 2) - IOW - to bypass the versioning from URW++ completely, and have Version field based on snapshot string, in a way: X.Y.Z == YYYY.MM.DD
Or do you some problem with this approach?
As I said, please do not invent the version on your own. Please apply the existing snapshot guidelines instead, i.e.: Version: 0
Release: 0.N.YYYYMMDD or Release: 0.N.YYYYMMDDgitHASH
Or even better, as you proposed in the other e-mail, Version: YYYYMMDD. This is much nicer for users, and actually easier for the maintainer of the package. David seems to ignore this option for some reason, but it really seems the best one.
(If upstream does something crazy and it is necessary to change the versioning scheme again in the future, adding an epoch is always an option.)
Zbyszek
On Thu, Dec 01, 2016 at 01:37:18AM +0000, Zbigniew Jędrzejewski-Szmek wrote:
On Wed, Nov 30, 2016 at 05:35:19PM +0100, Dominik 'Rathann' Mierzejewski wrote:
On Wednesday, 30 November 2016 at 16:07, David Kaspar [Dee'Kej] wrote:
On Wed, Nov 30, 2016 at 12:38 PM, Dominik 'Rathann' Mierzejewski < dominik@greysector.net> wrote:
[...]
I'm not sure where your Version == 1.0 comes from. If they're versioned
only by date now, then you have two options. Use Version: 0 in the new package in anticipation of upstream eventually reintroducing semantic versioning. Or, Version: YYYYMMDD. Admittedly, the latter looks nicer and upstream already said they'll stick to it.
The Version == 1.0 comes from the source code of the fonts themselves. Running 'grep "Version" *.afm' tells me that there are all files with Version == 1.0, except two of them (which have Version
If they don't have the same version then it doesn't make sense to use the version of *some* of them as base.
If you worry about upstream versioning sanity, then stick with Version: 0 and follow the snapshot versioning guidelines.
There's also one more option, and that is to base the package on upstream's git repository and the snapshot scheme, because we would be using snapshot string in the package name anyway. And it would also solve one more issue that upstream is not shipping license files in the archive. (I have already contacted to correct this.)
The exact location of the source doesn't matter too much as long as it's official and pristine. I agree it might be better to use the git repo directly since it contains both the licence indication and its full text.
Upstream has heard to my request and fixed it. ( http://bugs.ghostscript.com/show_bug.cgi?id=697390)
And yes, what Douhlas wrote is correct (about the 35 fonts), and I will have that noted in the %description section.
Anyway, since determining the Version field is still unclear, I think the most sense to me right now is to proceed with option 2) - IOW - to bypass the versioning from URW++ completely, and have Version field based on snapshot string, in a way: X.Y.Z == YYYY.MM.DD
Or do you some problem with this approach?
As I said, please do not invent the version on your own. Please apply the existing snapshot guidelines instead, i.e.: Version: 0
Release: 0.N.YYYYMMDD or Release: 0.N.YYYYMMDDgitHASH
Or even better, as you proposed in the other e-mail, Version: YYYYMMDD. This is much nicer for users, and actually easier for the maintainer of the package. David seems to ignore this option for some reason, but it really seems the best one.
(If upstream does something crazy and it is necessary to change the versioning scheme again in the future, adding an epoch is always an option.)
Since I guess the idea of this thread is to avoid using epoch, I think the better approach is to stick with the Dominik's suggestions. It is what gives the most flexibility for future changes.
Pierre
OK, I will proceed with Domininik's or Zbigniew's idea. Thanks all to your suggestions. :)
David Kaspar [Dee'Kej] *Associate Software Engineer* *Brno, Czech Republic*
RED HAT | TRIED. TESTED. TRUSTED. Every airline in the Fortune 500 relies on Red Hat. Find out why at Trusted | Red Hat http://www.redhat.com/en/about/trusted.
On Thu, Dec 1, 2016 at 9:13 AM, Pierre-Yves Chibon pingou@pingoured.fr wrote:
On Thu, Dec 01, 2016 at 01:37:18AM +0000, Zbigniew Jędrzejewski-Szmek wrote:
On Wed, Nov 30, 2016 at 05:35:19PM +0100, Dominik 'Rathann' Mierzejewski
wrote:
On Wednesday, 30 November 2016 at 16:07, David Kaspar [Dee'Kej] wrote:
On Wed, Nov 30, 2016 at 12:38 PM, Dominik 'Rathann' Mierzejewski < dominik@greysector.net> wrote:
[...]
I'm not sure where your Version == 1.0 comes from. If they're
versioned
only by date now, then you have two options. Use Version: 0 in the
new
package in anticipation of upstream eventually reintroducing
semantic
versioning. Or, Version: YYYYMMDD. Admittedly, the latter looks
nicer
and upstream already said they'll stick to it.
The Version == 1.0 comes from the source code of the fonts
themselves.
Running 'grep "Version" *.afm' tells me that there are all files with Version == 1.0, except two of them (which have Version
If they don't have the same version then it doesn't make sense to use the version of *some* of them as base.
If you worry about upstream versioning sanity, then stick with Version: 0 and follow the snapshot versioning guidelines.
There's also one more option, and that is to base the package on upstream's git repository and the snapshot scheme, because we would be using snapshot string in the package name anyway. And it would also solve one more issue that upstream is not shipping license files in the archive. (I have already contacted to
correct
this.)
The exact location of the source doesn't matter too much as long
as it's
official and pristine. I agree it might be better to use the git
repo
directly since it contains both the licence indication and its full text.
Upstream has heard to my request and fixed it. ( http://bugs.ghostscript.com/show_bug.cgi?id=697390)
And yes, what Douhlas wrote is correct (about the 35 fonts), and I
will
have that noted in the %description section.
Anyway, since determining the Version field is still unclear, I
think the
most sense to me right now is to proceed with option 2) - IOW - to
bypass
the versioning from URW++ completely, and have Version field based on snapshot string, in a way: X.Y.Z == YYYY.MM.DD
Or do you some problem with this approach?
As I said, please do not invent the version on your own. Please apply the existing snapshot guidelines instead, i.e.: Version: 0
Release: 0.N.YYYYMMDD or Release: 0.N.YYYYMMDDgitHASH
Or even better, as you proposed in the other e-mail, Version: YYYYMMDD. This is much nicer for users, and actually easier for the maintainer of the package. David seems to ignore this option for some reason, but it really seems the best one.
(If upstream does something crazy and it is necessary to change the versioning scheme again in the future, adding an epoch is always an option.)
Since I guess the idea of this thread is to avoid using epoch, I think the better approach is to stick with the Dominik's suggestions. It is what gives the most flexibility for future changes.
Pierre _______________________________________________ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-leave@lists.fedoraproject.org
-
David Kaspar [Dee'Kej] -
Dominik 'Rathann' Mierzejewski -
Douglas Kosovic -
Pierre-Yves Chibon -
Zbigniew Jędrzejewski-Szmek