Re: [Buildroot] [PATCH 1/1] package/lrzsz: fix build with musl

[Julien Olivain via buildroot <buildroot@buildroot.org>]

Hi Fiona,

On 21/07/2025 11:14, Fiona Klute wrote:
> Hi Julien,
> 
> thanks for the review!
> 
> Am 20.07.25 um 13:35 schrieb Julien Olivain:
>> On 21/06/2025 00:18, Fiona Klute via buildroot wrote:
>>> lib/long-options.c failed to compile with musl for the same reason
>>> 0002-lib-long-options.c-include-stdlib.h.patch was added to fix,
>>> exit() being undefined. The fix is the same as well: include 
>>> stdlib.h.
>> 
>> Before we merge this patch, could you add a comment in the upstream
>> pull request linked in the patch in [1] proposing your change?
>> 
>> I know lrzsz is an ancient package, and this fork did not received
>> commits in the past years, but we never know ;)
> 
> Sure, I've just done that:
> https://github.com/UweOhse/lrzsz/pull/4#issuecomment-3095797981

With that, I applied to master, thanks.

>> For the record, there was an alternate proposal in:
>> https://patchwork.ozlabs.org/project/buildroot/ 
>> patch/20250326170814.2393-1-uuidxx@163.com/
>> but since all the 3 libcs seems to support that, we can go with your 
>> proposal.
>> 
>>> Fixes: b6784a1f1f ("package/lrzsz: fix build with GCC >= 14.x")
>>> Signed-off-by: Fiona Klute <fiona.klute@gmx.de>
>>> ---
>>> This compiles with uClibc-ng, too (without even a warning, I didn't
>>> test at runtime), so I'm not sure what those "non-GNU C libraries"
>>> where things break are. There are similar comments all over the 
>>> getopt
>>> code, so with the age of the code maybe way outdated ones, or
>>> non-Linux platforms.
>> 
>> Finally, if you want to runtime test, there a simple runtime test:
>> https://gitlab.com/buildroot.org/buildroot/-/blob/master/support/ 
>> testing/tests/package/test_lrzsz.py
>> the long-opt will be a bit tested with the --help.
> 
> That test currently uses only a glibc build config. Do you mean I could 
> modify it locally to test with uClibc, or that I should add test 
> variants with musl and uClibc? The latter would certainly be good for 
> reliability, but also add a lot of build time to the tests (that's why 
> I'm hesitating).

I meant to modify the test locally, or sneak the to command sequence to 
quickly
reproduce those in a qemu build with the right config.

I agree that lrzsz does not really need more testing (one is already 
enough).
Autobuilders will cover the rest...

> Best regards,
> Fiona

Best regards,

Julien.
_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot