[bug#43591] [PATCH core-updates] gnu: glibc-final: Catch all cases of a glibc user not requesting 64-bit offsets and then using readdir.

[Ludovic Courtès]

Hi,

Danny Milosavljevic <dannym@scratchpost.org> skribis:

> On Tue, 29 Sep 2020 22:52:10 +0200
> Ludovic Courtès <ludo@gnu.org> wrote:
>
>> Marius Bakke <marius@gnu.org> skribis:
>> 
>> > Arguably running code for foreign architectures through QEMU binfmt is
>> > something of a hack.  Mandating that every package *must* be patched to
>> > support it seems user-hostile.  I'm more in favor of dropping it on the
>> > build farm, or just keep fixing things on a per-package basis.  
>> 
>> I’m fine with dropping things on the build farm; it’s just about
>> modifying machines-for-berlin.scm in maintenance.git.  Any takers?  :-)
>
> I don't know what "dropping things on the build farm" means in this context.
> Dropping what exactly?

Dropping emulated builds, or at least 32-bit emulated builds.  We just
need to remove build machines from the file above.

>> The above would override the default CFLAGS in Autoconf-generated
>> configure scripts (which is “-O2 -g”).  
>
> That is correct.  I'm currently working on v2 (testing a patchset already)
> and I totally forgot to add "-g -O2" the first time around.  Also, glibc
> itself must NOT have -D_FILE_OFFSET_BITS=64 (it makes sense not to, too).
>
>>So we’d have to be cautious.
>> But I think a global solution is preferable to adding
>> -D_FILE_OFFSET_BITS=64 to tens of packages.
>
> I agree.
>
> I still would like to see what actually changes--and I think with
> guix-data-services it should actually be possible to compare derivations
> before-and-after and find out which derivations of which packages changed
> at all because of the global -D_FILE_OFFSET_BITS=64.  I'd like some help
> using guix-data-services to find that out.  Otherwise a risk estimation
> cannot be done.

A change in gnu-build-system would change all the derivations.  I don’t
think the Data Service can help us here.

> Technically, if a package used direct assembly offsets (for some unfathomable
> reason), it could have an undetectable problem with the size change of off_t
> (and also struct dirent).  So examining the source code of the most essential
> packages manually is still good.  That's what I did in
> branch wip-file-offsets-64.

Yeah.

> I'm in the process of testing a patchset that globally sets
>
>   CFLAGS="-D_FILE_OFFSET_BITS=64 -g -O2"
>
> instead.

OK.

> That alone is not enough since there are a lot of non-autotools projects that
> just ignore the environment variable entirely--not to mention languages other
> than C.

Yeah…

I have mixed feelings: fixing packages one by one doesn’t sound great,
but OTOH setting the ‘CFLAGS’ environment variable globally can have
unexpected side effects in some cases (overriding package-specific
CFLAGS) and zero effects in other cases (for non-Autoconf packages or
badly-written ‘configure.ac’ files), both of which would be hard to
detect.

~~~

If we take a step back: what’s the problem?  We have a problem with
emulated 32-bit architectures on 64-bit architectures.  But that’s OK,
we can stop those emulations for now.  Then we have packages that do not
support large files; it’s not great but evidently we can live with it.
:-)  Ideally, we’d report it upstream when we encounter it.

So to me that hints at targeted fixes: fixing key packages like CMake
(roughly what you already did) where lack of large file support can be
problematic.

Thoughts?

Ludo’.