https://bugs.freedesktop.org/show_bug.cgi?id=94231

--- Comment #1 from Emil Velikov emil.l.velikov@gmail.com --- Hi Mike,

man makedev/major/minor does not mention anything about this header, although it does mention (the missing) #define _BSD_SOURCE. Does that get thinks working again?

Can you please send a patch (git send-email ideally) to the list. Also please quote the build error and any related information on the topic that you can find.

If you can do the same for mesa that'll be greatly appreciated.

Thank you

https://bugs.freedesktop.org/show_bug.cgi?id=94231

--- Comment #3 from Emil Velikov emil.l.velikov@gmail.com --- Are you sure about the removal of "include sys/sysmacros.h" ? Afaics include/sys/types.h includes the posix/sys/types.h with the latter still having the include.

Bth I'm not sure if things can explode with the unconditional include, imho it's better to keep things as documented.

I.e. if the _BSD_SOURCE define does not help, then either glibc needs to be fixed or the manuals should be updated (the glibc fold might have some suggestions) ?

https://bugs.freedesktop.org/show_bug.cgi?id=94231

--- Comment #5 from Emil Velikov emil.l.velikov@gmail.com --- (In reply to Daniel Stone from comment #4)

_BSD_SOURCE gets defined if you call AC_USE_SYSTEM_EXTENSIONS

Are you sure about this ? Looking at my autoconf 2.69 (/usr/share/autoconf/autoconf/specific.m4) and it sets the following

__EXTENSIONS__ _ALL_SOURCE _GNU_SOURCE _POSIX_PTHREAD_SEMANTICS _TANDEM_SOURCE

Had no idea about the different interpretation of the defines between GNU vs BSD libc(s). Thanks for that.

We can at check if the explicit define fixes things on Linux and ping the BSD people for more robust solution ?

https://bugs.freedesktop.org/show_bug.cgi?id=94231

--- Comment #7 from Emil Velikov emil.l.velikov@gmail.com --- Hmm... Gentoo devs decided to patch glibc in order to remove the include.

At the same time they lean that applications (25+ based on the incomplete list) should be fixed to use sys/sysmacros.h. There's no justification and the manual does not support this :-(

To make it even more strange the patch isn't alongside the glibc-2.23.ebuild in the repo. Am I looking at the wrong repo ? Can anyone share a link with discussion behind this decision ?

https://bugs.freedesktop.org/show_bug.cgi?id=94231

--- Comment #9 from Mike Lothian mike@fireburn.co.uk --- Created attachment 121891 --> https://bugs.freedesktop.org/attachment.cgi?id=121891&action=edit Glibc patches

Here's the patch tarball for your convenience

https://bugs.freedesktop.org/show_bug.cgi?id=94231

--- Comment #11 from Emil Velikov emil.l.velikov@gmail.com --- Thanks for the links Mike.

Not sure about others, but I'm not a happy bunny.

So there's a issue/bug in stdlib.h (or was it g++), that leads to the inclusion of sys/types.h. To 'fix' this lets break the documented behaviour (for those of us that read the manpages). Hmm ... wait what ?

Sorry but this sounds backwards. If we cannot squash the issue, it should be documented properly with list of workarounds (as pointed in the redhat bugzilla). Yes it sucks, big time. Yet we shouldn't break one thing, in order to workaround another issue, should we ?

If this workaround is needed/applicable with other libc providers like musl (on Linux at least), then there should be a documentation update alongside the deprecation. Followed by a lengthy period _before_ removing things ?

https://bugs.freedesktop.org/show_bug.cgi?id=94231

Mike Frysinger vapier@gentoo.org changed:

What |Removed |Added ---------------------------------------------------------------------------- CC| |vapier@gentoo.org

--- Comment #12 from Mike Frysinger vapier@gentoo.org --- it's really not that difficult to test for & include the sys/sysmacros.h header directly when you're using the non-standard functions they provide. this is the API under BSD systems, and where Linux C libs have already moved/are moving towards. *this* is the deprecation period.

upstream glibc will take a more measured approach including updating their docs, but realistically, no one is going to read those until they see breakage. the Gentoo changes are only in an opt-in dev-only build, so it too isn't hitting people randomly.

as the upstream thread shows, if you used the (long time documented) AC_HEADER_MAJOR macro in autoconf, then this wouldn't have impacted drm.

https://bugs.freedesktop.org/show_bug.cgi?id=94231

--- Comment #14 from Mike Lothian mike@fireburn.co.uk --- I'll get the patches created for libdrm, mesa & xorg-server, Mike F please can you provide the location of where the updated docs are and I'll include it in the commit message

Emil is there anything else you'd like? I'm going to mention the I provided above too

https://bugs.freedesktop.org/show_bug.cgi?id=94231

--- Comment #16 from Mike Frysinger vapier@gentoo.org --- (In reply to Emil Velikov from comment #13)

that is happening in upstream glibc. i'm not going to bother in Gentoo.

no one is "pointing fingers". we detected an error in libdrm in Gentoo, hence you guys got a patch. i don't know why you're upset about this.

https://bugs.freedesktop.org/show_bug.cgi?id=94231

--- Comment #18 from Matt Turner mattst88@gmail.com --- I guess you can think of this as the deprecation period.

Documentation has been updated:

https://github.com/mkerrisk/man-pages/commit/3350a86413d770198e11fe8df9a3cd5...

https://bugs.freedesktop.org/show_bug.cgi?id=94231

--- Comment #20 from Matt Turner mattst88@gmail.com --- (In reply to Emil Velikov from comment #19)

That was fast, only a few hours ago the commit landed.

No objections about having this in, although let's use AC_CHECK_HEADERS.

I'm wondering if we shouldn't give it a day or two before landing though. Obviously waiting for a man-pages release would be a serious overkill (afaict it will be within ~4 months).

Do you want to commit it... or do you want me to send it to the list?

https://bugs.freedesktop.org/show_bug.cgi?id=94231

--- Comment #22 from Mike Frysinger vapier@gentoo.org --- Created attachment 124642 --> https://bugs.freedesktop.org/attachment.cgi?id=124642&action=edit using AC_CHECK_HEADERS()

https://bugs.freedesktop.org/show_bug.cgi?id=94231

Emil Velikov emil.l.velikov@gmail.com changed:

What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |RESOLVED Resolution|--- |FIXED

--- Comment #24 from Emil Velikov emil.l.velikov@gmail.com --- Thanks for updating the man pages as well as the patch(es) Mike. Hope you and/or others will get to resolving libstdc++ + _GNU_SOURCE issue that inspired this.

I've pushed the AC_HEADER_MAJOR patch to master.