[gentoo-qa] QA confusion: Upstream qa issues -> failing to install

[gentoo-qa] QA confusion: Upstream qa issues -> failing to install

[Thilo Bangert]

[-- Attachment #1: Type: Text/Plain, Size: 1753 bytes --]

Hi guys,

trying to install openssh on the dev profile one is greated with this 
bugger:

 * QA Notice: Package has poor programming practices which may compile
 *            fine but exhibit random runtime failures.
 * closefromtest.c:46: warning: implicit declaration of function 
‘closefrom’

 * Please do not file a Gentoo bug and instead report the above QA
 * issues directly to the upstream developers of this software.
 * Homepage: http://www.openssh.org/
 * ERROR: net-misc/openssh-5.5_p1-r2 failed:
 *   install aborted due to poor programming practices shown above
 * 
 * Call stack:
 *   misc-functions.sh, line 847:  Called install_qa_check
 *   misc-functions.sh, line 539:  Called die
 * The specific snippet of code:
 *                              hasq stricter ${FEATURES} && die "install 
aborted due to" \
 * 
 * If you need support, post the output of 'emerge --info =net-
misc/openssh-5.5_p1-r2',
 * the complete build log and the output of 'emerge -pqv =net-
misc/openssh-5.5_p1-r2'.
 * The complete build log is located at '/var/tmp/portage/net-
misc/openssh-5.5_p1-r2/temp/build.log'.
 * The ebuild environment file is located at '/var/tmp/portage/net-
misc/openssh-5.5_p1-r2/temp/environment'.
 * S: '/var/tmp/portage/net-misc/openssh-5.5_p1-r2/work/openssh-5.5p1'

This looks really really weird. On the one hand we are saying: "Its a QA 
issue but upstream should fix it." followed by "It so broke, we wont let 
you install it."

Perhaps qa warnings that fail when FEATURES="stricter", should not point 
to upstream for the fix. Instead we should take responsibility for these 
and consequently allow these bugs to linger in our bugzilla.

What do you tink?
kind regards
Thilo

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-qa] QA confusion: Upstream qa issues -> failing to install

[Markos Chandras]

[-- Attachment #1: Type: text/plain, Size: 2788 bytes --]

On Wed, Aug 25, 2010 at 01:42:12PM +0200, Thilo Bangert wrote:
> Hi guys,
> 
> trying to install openssh on the dev profile one is greated with this 
> bugger:
> 
>  * QA Notice: Package has poor programming practices which may compile
>  *            fine but exhibit random runtime failures.
>  * closefromtest.c:46: warning: implicit declaration of function 
> ‘closefrom’
>
This is a rather common bug for 64bit arches

See relevant patches [1][2]

>  * Please do not file a Gentoo bug and instead report the above QA
>  * issues directly to the upstream developers of this software.
>  * Homepage: http://www.openssh.org/
>  * ERROR: net-misc/openssh-5.5_p1-r2 failed:
>  *   install aborted due to poor programming practices shown above
>  * 
>  * Call stack:
>  *   misc-functions.sh, line 847:  Called install_qa_check
>  *   misc-functions.sh, line 539:  Called die
>  * The specific snippet of code:
>  *                              hasq stricter ${FEATURES} && die "install 
> aborted due to" \
>  * 
>  * If you need support, post the output of 'emerge --info =net-
> misc/openssh-5.5_p1-r2',
>  * the complete build log and the output of 'emerge -pqv =net-
> misc/openssh-5.5_p1-r2'.
>  * The complete build log is located at '/var/tmp/portage/net-
> misc/openssh-5.5_p1-r2/temp/build.log'.
>  * The ebuild environment file is located at '/var/tmp/portage/net-
> misc/openssh-5.5_p1-r2/temp/environment'.
>  * S: '/var/tmp/portage/net-misc/openssh-5.5_p1-r2/work/openssh-5.5p1'
> 
> This looks really really weird. On the one hand we are saying: "Its a QA 
> issue but upstream should fix it." followed by "It so broke, we wont let 
> you install it."
This is not only related to stricter but also related to CFLAGS ( having -ggdb
and -Wall, produce same failures for me even if I don't use stricter )

Diego, any ideas why this happens? -Wall & -ggdb produce extra output on
warning that makes portage abort because of that? I always wondered why this
is happening.
> 
> Perhaps qa warnings that fail when FEATURES="stricter", should not point 
> to upstream for the fix. Instead we should take responsibility for these 
> and consequently allow these bugs to linger in our bugzilla.
> 
> What do you tink?
> kind regards
> Thilo

[1]:
http://sources.gentoo.org/cgi-bin/viewvc.cgi/gentoo-x86/net-analyzer/nagios-plugins/files/nagios-plugins-1.4.14-implicit-pointer-conversion.patch?view=log
[2]:
http://sources.gentoo.org/cgi-bin/viewvc.cgi/gentoo-x86/app-editors/xemacs/files/xemacs-21.4.22-implicit-pointer-conversion.patch?view=log

-- 
Markos Chandras (hwoarang)
Gentoo Linux Developer
Web: http://hwoarang.silverarrow.org
Key ID: 441AC410
Key FP: AAD0 8591 E3CD 445D 6411 3477 F7F7 1E8E 441A C410

[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-qa] QA confusion: Upstream qa issues -> failing to install

[Diego Elio Pettenò]

[-- Attachment #1: Type: text/plain, Size: 2293 bytes --]

Il giorno mer, 25/08/2010 alle 17.20 +0300, Markos Chandras ha scritto:
> On Wed, Aug 25, 2010 at 01:42:12PM +0200, Thilo Bangert wrote:
> >  * QA Notice: Package has poor programming practices which may compile
> >  *            fine but exhibit random runtime failures.
> >  * closefromtest.c:46: warning: implicit declaration of function 
> > ‘closefrom’
> >
> This is a rather common bug for 64bit arches

Actually no; the one common for 64bit arches is "implicit pointer
conversion" warning (which may be caused by an implicit declaration,
since implicit declarations default to int as return type!) that is
always fatal on 64-bit (it would crash the code as the returned pointer
would be crippled.

What this bug is about is rather the kind of errors that I've commented
on in [1], [2] and [3] which is that they _may_ be related to macros or
functions that are not declared/defined by the current set of library
dependencies, and would then lead to unresolved undefined symbols, and
thus, to runtime failure.

> This is not only related to stricter but also related to CFLAGS ( having -ggdb
> and -Wall, produce same failures for me even if I don't use stricter )
> 
> Diego, any ideas why this happens? -Wall & -ggdb produce extra output on
> warning that makes portage abort because of that? I always wondered why this
> is happening.

They are not issues for users to care much about, but if _you_ are
maintaining a package that have those warnings should probably go fix
them; upstream and in Gentoo; in my last blog post [4] I have noted that
we should really just use stricter for _our_ packages, not _all_
packages, as it makes it bothersome and lead to more people just
forgetting about stricter (I don't use it, for instance).

[1] http://blog.flameeyes.eu/2010/08/18/compounded-issues-in-glibc-2-12
[2] http://blog.flameeyes.eu/2010/08/19/fixed-in-overlay-read-not-fixed
[3] http://blog.flameeyes.eu/2010/02/03/ruby-ng-the-nth-fork
[4]
http://blog.flameeyes.eu/2010/08/24/gentoo-needs-you-a-few-things-that-would-definitely-be-useful

-- 
Diego Elio Pettenò — “Flameeyes”
http://blog.flameeyes.eu/

If you found a .asc file in this mail and know not what it is,
it's a GnuPG digital signature: http://www.gnupg.org/


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 490 bytes --]

Re: [gentoo-qa] QA confusion: Upstream qa issues -> failing to install

[Mike Frysinger]

[-- Attachment #1: Type: Text/Plain, Size: 655 bytes --]

On Wednesday, August 25, 2010 11:55:07 Diego Elio Pettenò wrote:
> What this bug is about is rather the kind of errors that I've commented
> on in [1], [2] and [3] which is that they _may_ be related to macros or
> functions that are not declared/defined by the current set of library
> dependencies, and would then lead to unresolved undefined symbols, and
> thus, to runtime failure.

or:
 - the return value is larger than an "int", and thus possibly truncated
 - the func requires 3 args, but user passes some other number, and compiler 
cannot flag it
 - the user passes args in the incorrect order and the compiler cant check it
-mike

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 836 bytes --]

Re: [gentoo-qa] QA confusion: Upstream qa issues -> failing to install

[Thilo Bangert]

[-- Attachment #1: Type: Text/Plain, Size: 1210 bytes --]

Mike Frysinger <vapier@gentoo.org> said:
> On Wednesday, August 25, 2010 11:55:07 Diego Elio Pettenò wrote:
> > What this bug is about is rather the kind of errors that I've
> > commented on in [1], [2] and [3] which is that they _may_ be related
> > to macros or functions that are not declared/defined by the current
> > set of library dependencies, and would then lead to unresolved
> > undefined symbols, and thus, to runtime failure.
> 
> or:
>  - the return value is larger than an "int", and thus possibly
> truncated - the func requires 3 args, but user passes some other
> number, and compiler cannot flag it
>  - the user passes args in the incorrect order and the compiler cant
> check it -mike

this is all very helpful info about the specific problem. we should 
probably collect this kind of information in a (wiki) page somewhere and 
point maintainers to it directly in the warning.

this would also help maintainers, when pushing patches upstream.

overall, i believe, that making the maintainers job easy, by providing him 
ample guidance and knowledge on how to fix specific issues will make it 
much more likely for him to fix the issue. this works for me anyway.


[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-qa] QA confusion: Upstream qa issues -> failing to install

[Markos Chandras]

[-- Attachment #1: Type: text/plain, Size: 1301 bytes --]

On Wed, Aug 25, 2010 at 05:55:07PM +0200, Diego Elio Pettenò wrote:
> Il giorno mer, 25/08/2010 alle 17.20 +0300, Markos Chandras ha scritto:
> > On Wed, Aug 25, 2010 at 01:42:12PM +0200, Thilo Bangert wrote:
> > >  * QA Notice: Package has poor programming practices which may compile
> > >  *            fine but exhibit random runtime failures.
> > >  * closefromtest.c:46: warning: implicit declaration of function 
> > > ‘closefrom’
> > >
> > This is a rather common bug for 64bit arches
> 
> Actually no; the one common for 64bit arches is "implicit pointer
> conversion" warning (which may be caused by an implicit declaration,
> since implicit declarations default to int as return type!) that is
> always fatal on 64-bit (it would crash the code as the returned pointer
> would be crippled.
> 
Diego thanks for the info. Can you please explain me why this is fatal only
with certain cflags combination? If I use -march=native -O2 -pipe the
compilation doesn't fail but if I add -ggdb -Wall the compilation fails. I
don't quite understand why so I would be grateful if you could enlighten me


-- 
Markos Chandras (hwoarang)
Gentoo Linux Developer
Web: http://hwoarang.silverarrow.org
Key ID: 441AC410
Key FP: AAD0 8591 E3CD 445D 6411 3477 F7F7 1E8E 441A C410

[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-qa] QA confusion: Upstream qa issues -> failing to install

[Diego Elio Pettenò]

Il giorno mer, 25/08/2010 alle 22.20 +0300, Markos Chandras ha scritto:
> 
> Diego thanks for the info. Can you please explain me why this is fatal
> only
> with certain cflags combination? If I use -march=native -O2 -pipe the
> compilation doesn't fail but if I add -ggdb -Wall the compilation
> fails. I
> don't quite understand why so I would be grateful if you could
> enlighten me
> 
> 
Simply the warning is not enabled by default, but only with -Wall or
-Wimplicit-declaration.

-- 
Diego Elio Pettenò — “Flameeyes”
http://blog.flameeyes.eu/

If you found a .asc file in this mail and know not what it is,
it's a GnuPG digital signature: http://www.gnupg.org/

Re: [gentoo-qa] QA confusion: Upstream qa issues -> failing to install

[Markos Chandras]

[-- Attachment #1: Type: text/plain, Size: 881 bytes --]

On Wed, Aug 25, 2010 at 11:51:35PM +0200, Diego Elio Pettenò wrote:
> Il giorno mer, 25/08/2010 alle 22.20 +0300, Markos Chandras ha scritto:
> > 
> > Diego thanks for the info. Can you please explain me why this is fatal
> > only
> > with certain cflags combination? If I use -march=native -O2 -pipe the
> > compilation doesn't fail but if I add -ggdb -Wall the compilation
> > fails. I
> > don't quite understand why so I would be grateful if you could
> > enlighten me
> > 
> > 
> Simply the warning is not enabled by default, but only with -Wall or
> -Wimplicit-declaration.
> 
Ah so the warning that pops up because of the -Wall, makes src_compile to die

Ok fair enough

Thanks :)

-- 
Markos Chandras (hwoarang)
Gentoo Linux Developer
Web: http://hwoarang.silverarrow.org
Key ID: 441AC410
Key FP: AAD0 8591 E3CD 445D 6411 3477 F7F7 1E8E 441A C410

[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-qa] QA confusion: Upstream qa issues -> failing to install

[Mark Loeser]

[-- Attachment #1: Type: text/plain, Size: 1064 bytes --]

Thilo Bangert <bangert@gentoo.org> said:
> trying to install openssh on the dev profile one is greated with this 
> bugger:
> 
>  * QA Notice: Package has poor programming practices which may compile
>  *            fine but exhibit random runtime failures.
>  * closefromtest.c:46: warning: implicit declaration of function 
> ‘closefrom’
> 
>  * Please do not file a Gentoo bug and instead report the above QA
>  * issues directly to the upstream developers of this software.


> This looks really really weird. On the one hand we are saying: "Its a QA 
> issue but upstream should fix it." followed by "It so broke, we wont let 
> you install it."


Personally I think we should be directing our users to our Bugzilla
always.  We really don't want to be pissing off upstreams if we put in a
patch that triggers one of these QA warnings and it ends up being our
problem and not theirs.

-- 
Mark Loeser
email         -   halcy0n AT gentoo DOT org
email         -   mark AT halcy0n DOT com
web           -   http://www.halcy0n.com

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-qa] QA confusion: Upstream qa issues -> failing to install

[Thilo Bangert]

[-- Attachment #1: Type: Text/Plain, Size: 1352 bytes --]

Mark Loeser <halcy0n@gentoo.org> said:
> Thilo Bangert <bangert@gentoo.org> said:
> > trying to install openssh on the dev profile one is greated with this
> > 
> > bugger:
> >  * QA Notice: Package has poor programming practices which may
> >  compile *            fine but exhibit random runtime failures.
> >  * closefromtest.c:46: warning: implicit declaration of function
> > 
> > ‘closefrom’
> > 
> >  * Please do not file a Gentoo bug and instead report the above QA
> >  * issues directly to the upstream developers of this software.
> > 
> > This looks really really weird. On the one hand we are saying: "Its a
> > QA issue but upstream should fix it." followed by "It so broke, we
> > wont let you install it."
> 
> Personally I think we should be directing our users to our Bugzilla
> always.  We really don't want to be pissing off upstreams if we put in
> a patch that triggers one of these QA warnings and it ends up being
> our problem and not theirs.

 i would agree, that all issues should be reported in our own bugzilla 
(also).

It would be really nice if we somehow could annotate the bug, in the 
ebuild, so that the warning already includes the bug number (or is 
silenced by it) in an effort to reduce the number of duplicates.

where are these qa checks implemented anyway?

thanks
Thilo

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-qa] QA confusion: Upstream qa issues -> failing to install

[Mark Loeser]

[-- Attachment #1: Type: text/plain, Size: 1826 bytes --]

Thilo Bangert <bangert@gentoo.org> said:
> Mark Loeser <halcy0n@gentoo.org> said:
> > Thilo Bangert <bangert@gentoo.org> said:
> > > trying to install openssh on the dev profile one is greated with this
> > > 
> > > bugger:
> > >  * QA Notice: Package has poor programming practices which may
> > >  compile *            fine but exhibit random runtime failures.
> > >  * closefromtest.c:46: warning: implicit declaration of function
> > > 
> > > ‘closefrom’
> > > 
> > >  * Please do not file a Gentoo bug and instead report the above QA
> > >  * issues directly to the upstream developers of this software.
> > > 
> > > This looks really really weird. On the one hand we are saying: "Its a
> > > QA issue but upstream should fix it." followed by "It so broke, we
> > > wont let you install it."
> > 
> > Personally I think we should be directing our users to our Bugzilla
> > always.  We really don't want to be pissing off upstreams if we put in
> > a patch that triggers one of these QA warnings and it ends up being
> > our problem and not theirs.
> 
>  i would agree, that all issues should be reported in our own bugzilla 
> (also).
> 
> It would be really nice if we somehow could annotate the bug, in the 
> ebuild, so that the warning already includes the bug number (or is 
> silenced by it) in an effort to reduce the number of duplicates.
> 
> where are these qa checks implemented anyway?

/usr/lib/portage/bin/misc-functions.sh

We should probably get a discussion going with the dev community as a
whole as to what we should change this text to, and why, so we don't
bother upstreams with our own possible problems.


-- 
Mark Loeser
email         -   halcy0n AT gentoo DOT org
email         -   mark AT halcy0n DOT com
web           -   http://www.halcy0n.com

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]