www/{*chromium,iridium}: Fix build after src snapshot 0c47b9c211

[jrm]

Summary

After src snapshot 0c47b9c211, the build began failing with:

../../third_party/nasm/include/compiler.h:249:21: error: static
declaration of 'mempcpy' follows non-static declaration

249 | static inline void *mempcpy(void *dst, const void *src, size_t n)
    |                     ^

/usr/include/string.h:71:8: note: previous declaration is here

71 | void    *(mempcpy)(void * __restrict, const void * __restrict, size_t);
   |           ^

• Restore a default definition of HAVE_MEMPCPY to indicate that mempcpy(3) is available. It has been part of FreeBSD since 13.1.

• Remove a patch used to detect mempcpy(3) in /usr/include/string.h. It is no longer necessary and does not detect the function after src snapshot 0c47b9c211.

Test Plan

Some tests are still running and will take days to complete on my build system. However, I think we are safe, given that the same changes have been applied to other ports that bundle nasm.

[jrm]

Also remove patch-third__party_nasm_BUILD.gn.

[jrm]

Tweak commit message.

[rnagy]

I'd rather much prefer third_party/nasm/BUILD.gn to be fixed. Also you are not adding HAVE_MEMPCPY , you are removing it.

[jrm]

In D46055#1050309, @rnagy wrote:

I'd rather much prefer third_party/nasm/BUILD.gn to be fixed.

I can rework the regex and restore patch-third__party_nasm_BUILD.gn, but it seems unnecessary. Could you elaborate on why it's still useful? We are testing for something with a certain answer because all our systems have supported mempcpy() since 13.1. Moreover, all our other bundled nasm (multimedia/ringrtc, devel/electron*, www/qt6-webengine) have now removed the patch. It would be nice to have all these ports consistent.

Also you are not adding HAVE_MEMPCPY , you are removing it.

The unpatched third_party/nasm/config/config-linux.h has #define HAVE_MEMPCPY 1. I'm removing a part of our patch that removed #define HAVE_MEMPCPY 1, i.e., restoring upstream's default of including #define HAVE_MEMPCPY 1.

[rnagy]

In D46055#1050502, @jrm wrote:

In D46055#1050309, @rnagy wrote:

I'd rather much prefer third_party/nasm/BUILD.gn to be fixed.

I can rework the regex and restore patch-third__party_nasm_BUILD.gn, but it seems unnecessary. Could you elaborate on why it's still useful? We are testing for something with a certain answer because all our systems have supported mempcpy() since 13.1. Moreover, all our other bundled nasm (multimedia/ringrtc, devel/electron*, www/qt6-webengine) have now removed the patch. It would be nice to have all these ports consistent.

Also you are not adding HAVE_MEMPCPY , you are removing it.

The unpatched third_party/nasm/config/config-linux.h has #define HAVE_MEMPCPY 1. I'm removing a part of our patch that removed #define HAVE_MEMPCPY 1, i.e., restoring upstream's default of including #define HAVE_MEMPCPY 1.

Because the patches of *chromium are shared between *BSDs, and not all of them have mempcpy so detecting its availability is necessary. I can have a look into this later this week if you want me to.

[jrm]

Understood. Here is another attempt, which fixes the build on FreeBSD main and restores the detection of mempcpy in string.h.