pkgsMusl.nginx: fix build

[superherointj]

pkgsMusl.nginx: fix build

Fixes:

error: #warning usage of non-standard #include <sys/cdefs.h> is deprecated [-Werror=cpp]

CC Musl: @yu-re-ka @alyssais @ulrikstrid
CC Nginx: @fpletz @RaitoBezarius @helsinki-systems

[alyssais]

We should strongly consider just ditching these compatibility headers and fixing stuff properly upstream, because I have a feeling they cause more build failures (like this one) than they prevent.

But let's worry about that after #229439.

[alyssais]

@ofborg build nginx pkgsMusl.nginx

[SuperSandro2000]

I think a very similar fix was already proposed half a year back in #263147 which was not accepted. Glad to see that we choose a churn friendly solution 😄

[superherointj]

Glad to see that we choose a churn friendly solution 😄

Hi Sandro,
As always, nice to see you here.
Thanks for the reference.
I'm not sure what you mean by that. I didn't have a better solution at the time.
I proposed what I found to be working. I understand that it might not be a perfect solution.
But I don't have perfection information. I don't really know C/C++ well [haven't programmed in that]. So I'm not comfortable in fixing at upstream level.
Also, time is limited, and there's priorities too. It's just too much to be done.
There is also an information imbalance between different people. Sometimes people say something like that is obvious, while that isn't for others.
Anyway, if someone has a better solution, I'm all for it.
Thanks for your understanding.

[Mic92]

Glad to see that we choose a churn friendly solution 😄

Hi Sandro, As always, nice to see you here. Thanks for the reference. I'm not sure what you mean by that. I didn't have a better solution at the time. I proposed what I found to be working. I understand that it might not be a perfect solution. But I don't have perfection information. I don't really know C/C++ well [haven't programmed in that]. So I'm not comfortable in fixing at upstream level. Also, time is limited, and there's priorities too. It's just too much to be done. There is also an information imbalance between different people. Sometimes people say something like that is obvious, while that isn't for others. Anyway, if someone has a better solution, I'm all for it. Thanks for your understanding.

I don't think he meant to criticize your approach, but instead was a bit disgruntled that his pull request was rejected. I would have probably fixed it in the same way. -Werror is better left to the projects itself, that can pin point a specific compiler version. However in distributions, you want to be able to upgrade compilers without having to patch half the ecosystem.

[SuperSandro2000]

But I don't have perfection information. I don't really know C/C++ well [haven't programmed in that]. So I'm not comfortable in fixing at upstream level.
Also, time is limited, and there's priorities too. It's just too much to be done.

Same, same for me. I totally didn't want to critique you and just say that I had the basically the same mindset a while ago and agree with the sentiment in this thread and that I am glad that I am not alone with this opinion.

[Ma27]

@SuperSandro2000 can you elaborate please? Seems like the PR was closed by the original author and it was your suggestion in review that got rejected.

[SuperSandro2000]

I think the original author and me had the same idea, to make musl maintenance easy and reduce churn with every update and compiler change but it got declined by the maintainer for reasons I do not follow. But everything alright, I don't think we need to change anything here codewise.