[devel] (gcc-gnat) repocop-lintian: bad-permissions-for-ali-file

[devel] (gcc-gnat) repocop-lintian: bad-permissions-for-ali-file

[Igor Vlasenko]

Господа, есть ругань на наши gcc-gnat.

Tag: bad-permissions-for-ali-file
Type: warning
Info: Ada Library Information (*.ali) files are required to be read-only
 (mode 0444) by GNAT. 
 .
 If at least one user can write the *.ali file, GNAT considers whether
 or not to recompile the corresponding source file.  Such recompilation
 would fail because normal users don't have write permission on the
 files.  Moreover, such recompilation would defeat the purpose of
 library packages, which provide *.a and *.so libraries to link against).

gcc3.4-gnat-3.4.5-alt7.x86_64
gcc4.1-gnat-4.1.2-alt2.x86_64
i386-mingw32msvc-gcc-gnat-3.4.5-alt2.x86_64

-- 

Dr. Igor Vlasenko
--------------------
Topology Department
Institute of Math
Kiev, Ukraine

Re: [devel] (gcc-gnat) repocop-lintian: bad-permissions-for-ali-file

[Dmitry V. Levin]

[-- Attachment #1: Type: text/plain, Size: 961 bytes --]

On Tue, May 06, 2008 at 10:31:31AM +0300, Igor Vlasenko wrote:
> Господа, есть ругань на наши gcc-gnat.
> 
> Tag: bad-permissions-for-ali-file
> Type: warning
> Info: Ada Library Information (*.ali) files are required to be read-only
>  (mode 0444) by GNAT. 

Надо поправить тест, поскольку права доступа 0644 вполне адекватны
приведённому ниже обоснованию.  Кроме того, на процессы с fsuid==0
эти ограничения прав доступа обычно не распространяются.

>  If at least one user can write the *.ali file, GNAT considers whether
>  or not to recompile the corresponding source file.  Such recompilation
>  would fail because normal users don't have write permission on the
>  files.  Moreover, such recompilation would defeat the purpose of
>  library packages, which provide *.a and *.so libraries to link against).
> 
> gcc3.4-gnat-3.4.5-alt7.x86_64
> gcc4.1-gnat-4.1.2-alt2.x86_64
> i386-mingw32msvc-gcc-gnat-3.4.5-alt2.x86_64


-- 
ldv

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]

Re: [devel] (gcc-gnat) repocop-lintian: bad-permissions-for-ali-file

[Igor Vlasenko]

On Tue, May 06, 2008 at 01:21:37PM +0400, Dmitry V. Levin wrote:
> Надо поправить тест, поскольку права доступа 0644 вполне адекватны
> приведённому ниже обоснованию.  Кроме того, на процессы с fsuid==0
> эти ограничения прав доступа обычно не распространяются.

Я вот не уверен. Вот что говорит народ:

http://osdir.com/ml/debian.devel.policy/2004-02/msg00027.html
ill Allombert <allomber@xxxxxxxxxxxxxxxxxx> writes:
> Why is it not sufficient for the file to be `write-protected' for the
> user running gnat instead of all users including root ? Or if it is, why
> when running as root cannot it add the compiled files in the same
> directory ? If it is not, why not just fix the test ?

The file must be 0444, not 0644.  The reason for this is within GNAT
itself; I think it has to do with portability (with Windows, OpenVMS).
If we would change this for Debian, we would become incompatible with
the other platforms and distributions where GNAT runs.

Одним словом, надо выяснить.

-- 

Dr. Igor Vlasenko
--------------------
Topology Department
Institute of Math
Kiev, Ukraine

Re: [devel] (gcc-gnat) repocop-lintian: bad-permissions-for-ali-file

[Alexey Tourbin]

[-- Attachment #1: Type: text/plain, Size: 738 bytes --]

On Tue, May 06, 2008 at 10:31:31AM +0300, Igor Vlasenko wrote:
> Господа, есть ругань на наши gcc-gnat.
> 
> Tag: bad-permissions-for-ali-file
> Type: warning
> Info: Ada Library Information (*.ali) files are required to be read-only
>  (mode 0444) by GNAT. 
>  .
>  If at least one user can write the *.ali file, GNAT considers whether
>  or not to recompile the corresponding source file.  Such recompilation
>  would fail because normal users don't have write permission on the
>  files.  Moreover, such recompilation would defeat the purpose of
>  library packages, which provide *.a and *.so libraries to link against).

Это какой-то бред сельских дедов.  Рут может писать в файлы с любыми
правами доступа, хоть 000.

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]

Re: [devel] (gcc-gnat) repocop-lintian: bad-permissions-for-ali-file

[Mikhail Gusarov]

[-- Attachment #1: Type: text/plain, Size: 952 bytes --]

Twas brillig at 19:18:29 06.05.2008 UTC+04 when Alexey Tourbin did gyre and gimble:

 >> Info: Ada Library Information (*.ali) files are required to be read-only
 >>  (mode 0444) by GNAT. 
 >>  .
 >>  If at least one user can write the *.ali file, GNAT considers whether
 >>  or not to recompile the corresponding source file.  Such recompilation
 >>  would fail because normal users don't have write permission on the
 >>  files.  Moreover, such recompilation would defeat the purpose of
 >>  library packages, which provide *.a and *.so libraries to link against).

 AT> Это какой-то бред сельских дедов.  Рут может писать в файлы с любыми
 AT> правами доступа, хоть 000.

Расширенное описание теста кривое. GNAT просто проверяет три w-бита.

-- 
JID: dottedmag@altlinux.org / dottedmag@jabber.dottedmag.net

[-- Attachment #2: Type: application/pgp-signature, Size: 196 bytes --]

Re: [devel] (gcc-gnat) repocop-lintian: bad-permissions-for-ali-file

[Alexey Tourbin]

[-- Attachment #1: Type: text/plain, Size: 1089 bytes --]

On Wed, May 07, 2008 at 01:08:23AM +0700, Mikhail Gusarov wrote:
> Twas brillig at 19:18:29 06.05.2008 UTC+04 when Alexey Tourbin did gyre and gimble:
> 
>  >> Info: Ada Library Information (*.ali) files are required to be read-only
>  >>  (mode 0444) by GNAT. 
>  >>  .
>  >>  If at least one user can write the *.ali file, GNAT considers whether
>  >>  or not to recompile the corresponding source file.  Such recompilation
>  >>  would fail because normal users don't have write permission on the
>  >>  files.  Moreover, such recompilation would defeat the purpose of
>  >>  library packages, which provide *.a and *.so libraries to link against).
> 
>  AT> Это какой-то бред сельских дедов.  Рут может писать в файлы с любыми
>  AT> правами доступа, хоть 000.
> 
> Расширенное описание теста кривое. GNAT просто проверяет три w-бита.

Так не стоит проверять, потому что реально возможность записи в файл
со стороны "at least one user" зависит от многих факторов (включая ACL,
опции монтирования и кое-что ещё).

Но можно проверять, например, access(file, W_OK).

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]

Re: [devel] (gcc-gnat) repocop-lintian: bad-permissions-for-ali-file

[Mikhail Gusarov]

[-- Attachment #1: Type: text/plain, Size: 598 bytes --]

Twas brillig at 22:49:22 06.05.2008 UTC+04 when Alexey Tourbin did gyre and gimble:

 AT> Так не стоит проверять, потому что реально возможность записи в файл со
 AT> стороны "at least one user" зависит от многих факторов (включая ACL, опции
 AT> монтирования и кое-что ещё).

Дыкть это и так ясно, но это уже к GNAT'овцам. Или к патченью GNAT'а нами.

-- 
JID: dottedmag@altlinux.org / dottedmag@jabber.dottedmag.net

[-- Attachment #2: Type: application/pgp-signature, Size: 196 bytes --]

Re: [devel] (gcc-gnat) repocop-lintian: bad-permissions-for-ali-file

[Dmitry V. Levin]

[-- Attachment #1: Type: text/plain, Size: 481 bytes --]

On Wed, May 07, 2008 at 03:04:37AM +0700, Mikhail Gusarov wrote:
> Twas brillig at 22:49:22 06.05.2008 UTC+04 when Alexey Tourbin did gyre and gimble:
> 
>  AT> Так не стоит проверять, потому что реально возможность записи в файл со
>  AT> стороны "at least one user" зависит от многих факторов (включая ACL, опции
>  AT> монтирования и кое-что ещё).
> 
> Дыкть это и так ясно, но это уже к GNAT'овцам. Или к патченью GNAT'а нами.

Повесьте баги там и сям.


-- 
ldv

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]