Bug#1024179: zlib breaks libcompress-raw-zlib-perl autopkgtest: 02zlib.t fails

[gregor herrmann]

On Wed, 16 Nov 2022 21:56:31 +0200, Niko Tyni wrote:

> From the log:
> 
>    not ok 2 - ZLIB_VERSION (1.2.11) matches Compress::Raw::Zlib::zlib_version
>    #     Failed test (t/compress/CompTestUtils.pm at line 61)
>    #          got: '1.2.11'
>    #     expected: '1.2.13'
>    # 
>    # The version of zlib.h does not match the version of libz
>    # 
>    # You have zlib.h version 1.2.11
>    #      and libz   version 1.2.13
>    # 
>    # You probably have two versions of zlib installed on your system.
>    # Try removing the one you don't want to use and rebuild.

Hu?

We have in d/rules:

override_dh_auto_test:
    TEST_SKIP_VERSION_CHECK=1 dh_auto_test

and that should be honoured in t/compress/CompTestUtils.pm.


But then, I also don't see this in
https://ci.debian.net/data/autopkgtest/testing/amd64/libc/libcompress-raw-zlib-perl/28323429/log.gz

But
t/01version.t ...... 
1..9
ok 1 - use Compress::Raw::Zlib;
ok 2 # skip TEST_SKIP_VERSION_CHECK is set


Is there a different log I'm missing?


But anyway, I guess this is the real problem:
    
> This is failing just because the runtime zlib version is not the same
> that the package was built with.
> 
>    not ok 126
>    #     Failed test (t/02zlib.t at line 498)
>    #          got: 1
>    #     expected: -3
>    [...]
>    not ok 134
>    #     Failed test (t/02zlib.t at line 532)
>    #          got: 1
>    #     expected: -3
>    
> These two trip on the same thing: apparently there's been a behaviour
> change in zlib 1.2.12. The test file has been adjusted for the change
> and has this comment:
> 
>   # Z_STREAM_END returned by 1.12.2, Z_DATA_ERROR for older zlib
> 
> but it's choosing the expected behaviour based on the build time zlib
> version rather than the runtime one.

Thanks for this analysis.

Some quick thoughts:
- we could play with zlib_version vs. ZLIB_VERSION (1.2.13 vs.
  1.2.11, according to t/000prereq.t) in t/02zlib.t
- we could skip these two tests with TEST_SKIP_VERSION_CHECK
- we could just do a binNMU
 
> I doubt the version skew causes any real problems but we can't really
> be sure about that. Applications could be choosing their behaviour based
> on ZLIB_VERSION. So just disarming the failing tests during autopkgtest
> and allowing version skew runs seems a bit risky.

*nod*
 
> The obvious alternative is to play it safe and add tight versioned
> dependencies on the zlib1g package, requiring rebuilds whenever there's
> a new zlib upstream version. That seems overkill to me.

Assuming this is not a recurring pattern, a one-time binNMU might be
not so bad …
 

Cheers,
gregor

-- 
 .''`.  https://info.comodo.priv.at -- Debian Developer https://www.debian.org
 : :' : OpenPGP fingerprint D1E1 316E 93A7 60A8 104D  85FA BB3A 6801 8649 AA06
 `. `'  Member VIBE!AT & SPI Inc. -- Supporter Free Software Foundation Europe
   `-   
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 963 bytes
Desc: Digital Signature
URL: <http://alioth-lists.debian.net/pipermail/pkg-perl-maintainers/attachments/20221116/3216404f/attachment-0001.sig>