Mass rebuild: git push --no-verify

[Jerry James]

Given the problems we had with font packages not being rebuilt in the
last mass rebuild, can we ensure that the mass rebuild script uses
"git push --no-verify" instead of plain "git push"?

See:
- https://bugzilla.redhat.com/show_bug.cgi?id=2233878
- https://lists.fedoraproject.org/archives/list/[e-mail]/...

[Sérgio Basto]

On Tue, 2024-01-16 at 14:50 -0700, Jerry James wrote:
> Given the problems we had with font packages not being rebuilt in the
> last mass rebuild, can we ensure that the mass rebuild script uses
> "git push --no-verify" instead of plain "git push"?
>
> See:
> https://bugzilla.redhat.com/show_bug.cgi?id=2233878
>
> https://lists.fedoraproject.org/archives/list/[e-mail]/...

You got mass rebuild script here [1] in massrebuildsinfo.py [2] you may
define what packages you are going to rebuild ,  in line 93 of mass-
rebuild.py [3] you got the list of packages that you go rebuild  
and since line 132 [4] you have the commands that will run . 
Is rpmdev-bumpspec that fails ?

[1]
https://pagure.io/releng/blob/main/f/scripts/mass-rebuild.py

[2]
https://pagure.io/releng/blob/main/f/scripts/massrebuildsinfo.py

[3] 
https://pagure.io/releng/blob/main/f/scripts/mass-rebuild.py#_93

[4]
https://pagure.io/releng/blob/main/f/scripts/mass-rebuild.py#_132
> -- 
> Jerry James
> http://www.jamezone.org/
> --
> _______________________________________________
> devel mailing list -- [e-mail]
> To unsubscribe send an email to [e-mail]
> Fedora Code of Conduct:
> https://docs.fedoraproject.org/en-US/project/code-of-conduct/
> List Guidelines:
> https://fedoraproject.org/wiki/Mailing_list_guidelines
> List Archives:
> https://lists.fedoraproject.org/archives/list/[e-mail]
> Do not reply to spam, report it:
> https://pagure.io/fedora-infrastructure/new_issue

[Jerry James]

On Tue, Jan 16, 2024 at 3:32 PM Sérgio Basto  wrote:
> You got mass rebuild script here [1] in massrebuildsinfo.py [2] you may
> define what packages you are going to rebuild ,  in line 93 of mass-
> rebuild.py [3] you got the list of packages that you go rebuild
> and since line 132 [4] you have the commands that will run .
> Is rpmdev-bumpspec that fails ?

Thank you for the pointer, Sérgio.  I have opened
https://pagure.io/releng/pull-request/11897.

It is not rpmdev-bumpspec that fails.  That works just fine.  But any
package that uses the %{fontpkgname1}, %{fontpkgname2}, ... macros
requires --no-verify today when pushing to git, due to the referenced
bug.  That's merely annoying for a human, but fatal to an automated
build script that doesn't pass --noverify.

[Sérgio Basto]

On Thu, 2024-01-18 at 09:13 -0700, Jerry James wrote:
> On Tue, Jan 16, 2024 at 3:32 PM Sérgio Basto 
> wrote:
> > You got mass rebuild script here [1] in massrebuildsinfo.py [2] you
> > may
> > define what packages you are going to rebuild ,  in line 93 of
> > mass-
> > rebuild.py [3] you got the list of packages that you go rebuild
> > and since line 132 [4] you have the commands that will run .
> > Is rpmdev-bumpspec that fails ?
>
> Thank you for the pointer, Sérgio.  I have opened
> https://pagure.io/releng/pull-request/11897.
>
> It is not rpmdev-bumpspec that fails.  That works just fine.  But any
> package that uses the %{fontpkgname1}, %{fontpkgname2}, ... macros
> requires --no-verify today when pushing to git, due to the referenced
> bug.  That's merely annoying for a human, but fatal to an automated
> build script that doesn't pass --noverify.

ok, is git push that fails .
As an idea , you may do "no verify" build option like noautobuild, in
line 143 of mass-rebuild.py [1] we have "Check for a noautobuild file"
to optout to skip mass rebuild, we may add an option in similar way for
noverify push, for example file "noverifybuild" .

Kodi use noautobuild (and it works :) ) :
https://github.com/rpmfusion/kodi/

[1]
https://pagure.io/releng/blob/main/f/scripts/mass-rebuild.py#_143
> -- 
> Jerry James
> http://www.jamezone.org/
> --
> _______________________________________________
> devel mailing list -- [e-mail]
> To unsubscribe send an email to [e-mail]
> Fedora Code of Conduct:
> https://docs.fedoraproject.org/en-US/project/code-of-conduct/
> List Guidelines:
> https://fedoraproject.org/wiki/Mailing_list_guidelines
> List Archives:
> https://lists.fedoraproject.org/archives/list/[e-mail]
> Do not reply to spam, report it:
> https://pagure.io/fedora-infrastructure/new_issue

[Michael J Gruber]

Am Do., 18. Jan. 2024 um 17:14 Uhr schrieb Jerry James :
> On Tue, Jan 16, 2024 at 3:32 PM Sérgio Basto  wrote:
> > You got mass rebuild script here [1] in massrebuildsinfo.py [2] you may
> > define what packages you are going to rebuild ,  in line 93 of mass-
> > rebuild.py [3] you got the list of packages that you go rebuild
> > and since line 132 [4] you have the commands that will run .
> > Is rpmdev-bumpspec that fails ?
>
> Thank you for the pointer, Sérgio.  I have opened
> https://pagure.io/releng/pull-request/11897.
>
> It is not rpmdev-bumpspec that fails.  That works just fine.  But any
> package that uses the %{fontpkgname1}, %{fontpkgname2}, ... macros
> requires --no-verify today when pushing to git, due to the referenced
> bug.  That's merely annoying for a human, but fatal to an automated
> build script that doesn't pass --noverify.

No, that's not the case. Please re-read the bug.

For *some* of the packages among those which use these macros (but
definitely not all), the python bindings to rpm do not expand that
macro whereas rpm itself does.

Since `spectool` uses the bindings, its check for the presence of all
source files fails if the source macro contains an unexpanded macro.
The packages build fine.

Whether it's the bindings' fault, or the spec files', or the
rpm-font-macros' is unclear at this point. In any case the check
*wrongly* indicates FTBFS.

Michael

[kevin]

On Tue, Jan 16, 2024 at 02:50:43PM -0700, Jerry James wrote:
> Given the problems we had with font packages not being rebuilt in the
> last mass rebuild, can we ensure that the mass rebuild script uses
> "git push --no-verify" instead of plain "git push"?
>
> See:
> https://bugzilla.redhat.com/show_bug.cgi?id=2233878
> https://lists.fedoraproject.org/archives/list/[e-mail]/...

I suppose this might be a good idea... I'd be afraid of what it might
break, while fixing the fonts packages though. But of course if it was
completely broken it would fail after that anyhow...

kevin

[Jerry James]

On Wed, Jan 17, 2024 at 3:48 PM kevin  wrote:
> I suppose this might be a good idea... I'd be afraid of what it might
> break, while fixing the fonts packages though. But of course if it was
> completely broken it would fail after that anyhow...

That's exactly my thinking.  The package is either broken already or
it is not.  If it is, then allowing a mass rebuild push won't change
the situation.  If it is not, then skipping verification has no
effect.

[Tomas Hrcka]

On Tue, Jan 16, 2024 at 10:52 PM Jerry James  wrote:
> Given the problems we had with font packages not being rebuilt in the
> last mass rebuild, can we ensure that the mass rebuild script uses
> "git push --no-verify" instead of plain "git push"?

This is not a good idea. Because of a few packages that are not rebuilding
we would disable the hook for every push the script does.
> See:
> https://bugzilla.redhat.com/show_bug.cgi?id=2233878
> https://lists.fedoraproject.org/archives/list/[e-mail]/...
>
> Jerry James
> http://www.jamezone.org/
> --
> _______________________________________________
> devel mailing list -- [e-mail]
> To unsubscribe send an email to [e-mail]
> Fedora Code of Conduct:
> https://docs.fedoraproject.org/en-US/project/code-of-conduct/
> List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
> List Archives:
> https://lists.fedoraproject.org/archives/list/[e-mail]
> Do not reply to spam, report it:
> https://pagure.io/fedora-infrastructure/new_issue

[Jerry James]

On Thu, Jan 18, 2024 at 4:50 AM Tomas Hrcka  wrote:
> This is not a good idea. Because of a few packages that are not rebuilding we would disable the hook for every push the script does.

My thinking is that the hook is not useful for automated build scripts
anyway, so disabling it doesn't hurt.  See my previous replies in this
thread.

[kevin]

On Thu, Jan 18, 2024 at 09:15:18AM -0700, Jerry James wrote:
> On Thu, Jan 18, 2024 at 4:50 AM Tomas Hrcka  wrote:
> > This is not a good idea. Because of a few packages that are not rebuilding we would disable the hook for every push the script does.
>
> My thinking is that the hook is not useful for automated build scripts
> anyway, so disabling it doesn't hurt.  See my previous replies in this
> thread.

yeah...

* Current setup with the hook:
  Bunch of fonts don't even get touched by the mass rebuild
  Some other packages with problems caught by the hook also are
completely missed.

* With --no-verify:
  Bunch of fonts do rebuild in the mass rebuild
  Some packages with problems get a mass rebuild commit at least and
attempted build. It might fail, but it's at least tried.

The mass rebuild is only doing a bump/rebuild. There's no reason it
should ever cause something that be caught by the hook, and if it did,
it would be better for it to do the commit anyhow and cause a failed
build. IMHO.

So, I think we should run with --no-verify.

kevin

[Björn Persson]

kevin wrote:
> The mass rebuild is only doing a bump/rebuild. There's no reason it
> should ever cause something that be caught by the hook, and if it did,
> it would be better for it to do the commit anyhow and cause a failed
> build. IMHO.

If, hypothetically, a defect in the mass-rebuild script would corrupt
thousands of spec files, how easy would it be to write a mass-revert
script to repair the damage? The mass-revert script shouldn't just
revert the latest commit in every package, because the corruption might
not have happened in every package, and some might have been reverted
manually in the meantime. The mass-revert script would need to verify
that it reverts only commits done by the defective mass-rebuild script.

If that's nontrivial to get right, then it seems to me that there is
value in a hook that validates changes made by a script.

Björn Persson

[kevin]

On Thu, Jan 18, 2024 at 08:24:38PM +0100, Björn Persson wrote:
> If, hypothetically, a defect in the mass-rebuild script would corrupt
> thousands of spec files, how easy would it be to write a mass-revert
> script to repair the damage? The mass-revert script shouldn't just
> revert the latest commit in every package, because the corruption might
> not have happened in every package, and some might have been reverted
> manually in the meantime. The mass-revert script would need to verify
> that it reverts only commits done by the defective mass-rebuild script.
>
> If that's nontrivial to get right, then it seems to me that there is
> value in a hook that validates changes made by a script.

That seems pretty hypothetical.

The pre-push check simply looks at the sources/patches defineed in the
spec and checks them against the sources file. The mass rebuild script
only uses rpmdev-bumpspec, which should only change the release and add
a changelog entry (or even less if the spec is using rpmautospec).

Should these font packages get fixed? Absolutely. 
But I think doing no-verify helps us because we will track more of those
packages that were simply skipped. I think it's better to not skip them
and have them fail than ignore them.

kevin

[Leslie Satenstein]

AgreeThank you.

See below

Leslie Satenstein

On Thursday, January 18, 2024 at 01:35:13 p.m. EST, kevin  wrote:

On Thu, Jan 18, 2024 at 09:15:18AM -0700, Jerry James wrote:
> On Thu, Jan 18, 2024 at 4:50 AM Tomas Hrcka  wrote:
> > This is not a good idea. Because of a few packages that are not rebuilding we would disable the hook for every push the script does.
>
> My thinking is that the hook is not useful for automated build scripts
> anyway, so disabling it doesn't hurt.  See my previous replies in this
> thread.

yeah...

* Current setup with the hook:
  Bunch of fonts don't even get touched by the mass rebuild
  Some other packages with problems caught by the hook also are
completely missed.

* With --no-verify:
  Bunch of fonts do rebuild in the mass rebuild
  Some packages with problems get a mass rebuild commit at least and
attempted build. It might fail, but it's at least tried.

The mass rebuild is only doing a bump/rebuild. There's no reason it
should ever cause something that be caught by the hook, and if it did,
it would be better for it to do the commit anyhow and cause a failed
build. IMHO.

So, I think we should run with --no-verify.

kevin
--
_______________________________________________
devel mailing list -- [e-mail]
To unsubscribe send an email to [e-mail]
Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/
List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
List Archives: https://lists.fedoraproject.org/archives/list/[e-mail]
Do not reply to spam, report it: https://pagure.io/fedora-infrastructure/new_issue