Rebuilding tesseract dependents in F42 for soversion change

[Michel Lind]

Dear all,

This PR: https://src.fedoraproject.org/rpms/tesseract/pull-request/6
was recently merged and built: https://bodhi.fedoraproject.org/updates/FEDORA-2024-41b92bf172

It backports a PR that's merged upstream to change the Tesseract
SOVERSION from maj.min.patch to only maj.min (Tesseract follows SemVer
and already uses maj.min for its Windows build, so the Linux versioning
was just a misconfiguration). So in the future, patch releases can be
built without having to rebuild dependents!

This also splits off tesseract-libs into a separate subpackage, which
will make it easier in the future to have compat versions of tesseract
since the libraries could coexist.

Unfortunately I omitted to mention that tesseract dependencies have to
be rebuilt - so we have to move slightly faster than usual to rebuild
the dependents.

This will be done in f42-build-side-96586 that's already created for
updating ffmpeg (one of those packages that need tesseract).

Best regards,

[Michael J Gruber]

Am Mi., 25. Sept. 2024 um 20:49 Uhr schrieb Michel Lind
:
> Dear all,
>
> This PR: https://src.fedoraproject.org/rpms/tesseract/pull-request/6
> was recently merged and built: https://bodhi.fedoraproject.org/updates/FEDORA-2024-41b92bf172
>
> It backports a PR that's merged upstream to change the Tesseract
> SOVERSION from maj.min.patch to only maj.min (Tesseract follows SemVer
> and already uses maj.min for its Windows build, so the Linux versioning
> was just a misconfiguration). So in the future, patch releases can be
> built without having to rebuild dependents!
>
> This also splits off tesseract-libs into a separate subpackage, which
> will make it easier in the future to have compat versions of tesseract
> since the libraries could coexist.
>
> Unfortunately I omitted to mention that tesseract dependencies have to
> be rebuilt - so we have to move slightly faster than usual to rebuild
> the dependents.
>
> This will be done in f42-build-side-96586 that's already created for
> updating ffmpeg (one of those packages that need tesseract).
>
> Best regards,

Hi Michel

Thanks for the heads-up and your work.

Are you planning to build recursive dependencies, or should we take
care of that?

I maintain mupdf and had been holding back a version update so far. It
would make sense to build this into your side-tag before you or I
rebuild mupdf's dependencies there.

Cheers
Michael

[Neal Gompa]

On Wed, Sep 25, 2024 at 2:49 PM Michel Lind  wrote:
> Dear all,
>
> This PR: https://src.fedoraproject.org/rpms/tesseract/pull-request/6
> was recently merged and built: https://bodhi.fedoraproject.org/updates/FEDORA-2024-41b92bf172
>
> It backports a PR that's merged upstream to change the Tesseract
> SOVERSION from maj.min.patch to only maj.min (Tesseract follows SemVer
> and already uses maj.min for its Windows build, so the Linux versioning
> was just a misconfiguration). So in the future, patch releases can be
> built without having to rebuild dependents!
>
> This also splits off tesseract-libs into a separate subpackage, which
> will make it easier in the future to have compat versions of tesseract
> since the libraries could coexist.
>
> Unfortunately I omitted to mention that tesseract dependencies have to
> be rebuilt - so we have to move slightly faster than usual to rebuild
> the dependents.
>
> This will be done in f42-build-side-96586 that's already created for
> updating ffmpeg (one of those packages that need tesseract).

I attempted a rebuild to deal with a bodhi error where it wouldn't let
me land it with the NVR reserved by another obsolete update, but I'm
getting bizarre failures for the mingw builds. Can someone help,
please?

Here's the failed build task:
https://koji.fedoraproject.org/koji/taskinfo?taskID=124237852

It seems like something has gone wrong with trying to use the C++
standard library in MinGW somehow?

[Sandro Mani]

On 30.09.24 23:35, Neal Gompa wrote:
> On Wed, Sep 25, 2024 at 2:49 PM Michel Lind  wrote:
> > Dear all,
> >
> > This PR: https://src.fedoraproject.org/rpms/tesseract/pull-request/6
> > was recently merged and built: https://bodhi.fedoraproject.org/updates/FEDORA-2024-41b92bf172
> >
> > It backports a PR that's merged upstream to change the Tesseract
> > SOVERSION from maj.min.patch to only maj.min (Tesseract follows SemVer
> > and already uses maj.min for its Windows build, so the Linux versioning
> > was just a misconfiguration). So in the future, patch releases can be
> > built without having to rebuild dependents!
> >
> > This also splits off tesseract-libs into a separate subpackage, which
> > will make it easier in the future to have compat versions of tesseract
> > since the libraries could coexist.
> >
> > Unfortunately I omitted to mention that tesseract dependencies have to
> > be rebuilt - so we have to move slightly faster than usual to rebuild
> > the dependents.
> >
> > This will be done in f42-build-side-96586 that's already created for
> > updating ffmpeg (one of those packages that need tesseract).
>
> I attempted a rebuild to deal with a bodhi error where it wouldn't let
> me land it with the NVR reserved by another obsolete update, but I'm
> getting bizarre failures for the mingw builds. Can someone help,
> please?
>
> Here's the failed build task:
> https://koji.fedoraproject.org/koji/taskinfo?taskID=124237852
>
> It seems like something has gone wrong with trying to use the C++
> standard library in MinGW somehow?

This is most likely due to the fact that mingw-crt-12.0.0 for the 
mingw32 and mingw64 variants was initially built with the incorrect 
default msvcrt. This has since been fixed, but mingw-gcc also needs to 
be rebuilt. I'm taking care of this now.

Apologies for the trouble.

Sandro

[Michael J Gruber]

Am Di., 1. Okt. 2024 um 00:07 Uhr schrieb Sandro Mani :
> On 30.09.24 23:35, Neal Gompa wrote:
> > On Wed, Sep 25, 2024 at 2:49 PM Michel Lind  wrote:
> > > Dear all,

...
> > > Unfortunately I omitted to mention that tesseract dependencies have to
> > > be rebuilt - so we have to move slightly faster than usual to rebuild
> > > the dependents.
> > >
> > > This will be done in f42-build-side-96586 that's already created for
> > > updating ffmpeg (one of those packages that need tesseract).
> >
> > I attempted a rebuild to deal with a bodhi error where it wouldn't let
> > me land it with the NVR reserved by another obsolete update, but I'm
> > getting bizarre failures for the mingw builds. Can someone help,
> > please?
> >
> > Here's the failed build task:
> > https://koji.fedoraproject.org/koji/taskinfo?taskID=124237852
> >
> > It seems like something has gone wrong with trying to use the C++
> > standard library in MinGW somehow?
>
> This is most likely due to the fact that mingw-crt-12.0.0 for the
> mingw32 and mingw64 variants was initially built with the incorrect
> default msvcrt. This has since been fixed, but mingw-gcc also needs to
> be rebuilt. I'm taking care of this now.

I'd like to reiterate my question about recursive dependencies:

│ Are you planning to build recursive dependencies, or should we take
│ care of that?
│
│ I maintain mupdf and had been holding back a version update so far. It
│ would make sense to build this into your side-tag before you or I
│ rebuild mupdf's dependencies there.

zathura-pudf-mupdf, python-PyMuPDF and qpdfview are affected by a
mupdf update which changes soname. They *should not break* due to the
side-tag rebuild which keeps soname. So it's more about timing and the
preference to separate or combine updates (for rawhide; for release
I'd combine).

Michael

[Neal Gompa]

On Tue, Oct 1, 2024 at 4:55 AM Michael J Gruber  wrote:
> Am Di., 1. Okt. 2024 um 00:07 Uhr schrieb Sandro Mani :
> > On 30.09.24 23:35, Neal Gompa wrote:
> > > On Wed, Sep 25, 2024 at 2:49 PM Michel Lind  wrote:
> > > > Dear all,
>
> ...
> > > > Unfortunately I omitted to mention that tesseract dependencies have to
> > > > be rebuilt - so we have to move slightly faster than usual to rebuild
> > > > the dependents.
> > > >
> > > > This will be done in f42-build-side-96586 that's already created for
> > > > updating ffmpeg (one of those packages that need tesseract).
> > >
> > > I attempted a rebuild to deal with a bodhi error where it wouldn't let
> > > me land it with the NVR reserved by another obsolete update, but I'm
> > > getting bizarre failures for the mingw builds. Can someone help,
> > > please?
> > >
> > > Here's the failed build task:
> > > https://koji.fedoraproject.org/koji/taskinfo?taskID=124237852
> > >
> > > It seems like something has gone wrong with trying to use the C++
> > > standard library in MinGW somehow?
> >
> > This is most likely due to the fact that mingw-crt-12.0.0 for the
> > mingw32 and mingw64 variants was initially built with the incorrect
> > default msvcrt. This has since been fixed, but mingw-gcc also needs to
> > be rebuilt. I'm taking care of this now.
>
> I'd like to reiterate my question about recursive dependencies:
>
> │ Are you planning to build recursive dependencies, or should we take
> │ care of that?
> │
> │ I maintain mupdf and had been holding back a version update so far. It
> │ would make sense to build this into your side-tag before you or I
> │ rebuild mupdf's dependencies there.
>
> zathura-pudf-mupdf, python-PyMuPDF and qpdfview are affected by a
> mupdf update which changes soname. They *should not break* due to the
> side-tag rebuild which keeps soname. So it's more about timing and the
> preference to separate or combine updates (for rawhide; for release
> I'd combine).

You should probably take care of it now in Rawhide, but there will be
a side-tag opened for F41 to bring back this there (as well as
ffmpeg7) that you may want to put that into so we can ship it all at
once.

[Sandro Mani]

On 01.10.24 00:06, Sandro Mani wrote:
> On 30.09.24 23:35, Neal Gompa wrote:
> > On Wed, Sep 25, 2024 at 2:49 PM Michel Lind  
> > wrote:
> > > Dear all,
> > >
> > > This PR: https://src.fedoraproject.org/rpms/tesseract/pull-request/6
> > > was recently merged and built: 
> > > https://bodhi.fedoraproject.org/updates/FEDORA-2024-41b92bf172
> > >
> > > It backports a PR that's merged upstream to change the Tesseract
> > > SOVERSION from maj.min.patch to only maj.min (Tesseract follows SemVer
> > > and already uses maj.min for its Windows build, so the Linux versioning
> > > was just a misconfiguration). So in the future, patch releases can be
> > > built without having to rebuild dependents!
> > >
> > > This also splits off tesseract-libs into a separate subpackage, which
> > > will make it easier in the future to have compat versions of tesseract
> > > since the libraries could coexist.
> > >
> > > Unfortunately I omitted to mention that tesseract dependencies have to
> > > be rebuilt - so we have to move slightly faster than usual to rebuild
> > > the dependents.
> > >
> > > This will be done in f42-build-side-96586 that's already created for
> > > updating ffmpeg (one of those packages that need tesseract).
> >
> > I attempted a rebuild to deal with a bodhi error where it wouldn't let
> > me land it with the NVR reserved by another obsolete update, but I'm
> > getting bizarre failures for the mingw builds. Can someone help,
> > please?
> >
> > Here's the failed build task:
> > https://koji.fedoraproject.org/koji/taskinfo?taskID=124237852
> >
> > It seems like something has gone wrong with trying to use the C++
> > standard library in MinGW somehow?
>
> This is most likely due to the fact that mingw-crt-12.0.0 for the 
> mingw32 and mingw64 variants was initially built with the incorrect 
> default msvcrt. This has since been fixed, but mingw-gcc also needs to 
> be rebuilt. I'm taking care of this now.
>
> Apologies for the trouble.

This should now be ok, a scratch build of tesseract with the rebuilt 
mingw-gcc succeeded.

Sandro

[Ron Yorston]

Sandro Mani  wrote:
> This is most likely due to the fact that mingw-crt-12.0.0 for the 
> mingw32 and mingw64 variants was initially built with the incorrect 
> default msvcrt. This has since been fixed, but mingw-gcc also needs to 
> be rebuilt. I'm taking care of this now.

Is it also necessary to rebuild other mingw32/mingw64 packages that
were built with the faulty mingw-crt-12.0.0?  Both in F41 and F42.

Ron