Bug 306533 – dev-libs/nspr: 'ipv6' useflag is probably pointless on both linux and freebsd

• Product: Gentoo Linux
• Status: RESOLVED FIXED

[Description Rafał Mużyło 2010-02-23 15:24:28 UTC]

...at least it seems that at most it will cause redefinition warnings.
Search the sources for '_PR_INET6' on linux, it's defines depending
on glibc version, on freebsd, depending on freebsd version.
See respective headers in mozilla/nspr/pr/include/md.

[Comment 1 Jory A. Pratt 2010-02-24 00:28:59 UTC]

It could be concider an automagic build, not the way we do things in gentoo.

[Comment 2 Rafał Mużyło 2010-02-24 00:38:56 UTC]

You've missed my point - it was: currently that useflag does nothing
neither on linux nor freebsd.

[Comment 3 Nirbheek Chauhan (RETIRED) 2010-02-28 07:16:49 UTC]

(In reply to comment #2)
> You've missed my point - it was: currently that useflag does nothing
> neither on linux nor freebsd.
> 

Yes, we know. It's there for two reasons:

a) So people don't file bugs about it being "automagic"
b) Future-proof; so that if upstream adds ipv6 code, we don't miss it and break setups all over the place.

[Comment 4 Rafał Mużyło 2010-02-28 19:14:14 UTC]

Could one of the devs PLEASE look at the source of nspr first ?

Right now, it wouldn't be automagic, even if the useflag was
dropped. See the response in https://bugzilla.mozilla.org/show_bug.cgi?id=545036#c14 .

Upstream does not support '--enable-ipv6' on Linux and FreeBSD
and it seems it has no plans to support it in the future.

I.e. for linux, there's a block:
#if (__GLIBC__ > 2) || (__GLIBC__ == 2 && __GLIBC_MINOR__ >= 1)
#define _PR_INET6
#define _PR_HAVE_INET_NTOP
#define _PR_HAVE_GETHOSTBYNAME2
#define _PR_HAVE_GETADDRINFO
#define _PR_INET6_PROBE
#endif

so, on a current glibc Gentoo it's *already* *always* enabled.
And '--disable-ipv6' will do nothing, as the only thing
'--enable-ipv6' does is
USE_IPV6=1
which seems to be aimed only at following make block:
ifeq ($(USE_IPV6),1)
DEFINES += -D_PR_INET6
endif

However, on a recent glibc system that's already defined.
Perhaps it would matter on a uClibc systems, but I don't think
mozilla upstream supports those.

[Comment 5 Jory A. Pratt 2010-02-28 19:31:41 UTC]

(In reply to comment #4)
> However, on a recent glibc system that's already defined.
> Perhaps it would matter on a uClibc systems, but I don't think
> mozilla upstream supports those.
> 

Does not matter if they support them or not, we do support them here in gentoo please do not reopen this bug.

[Comment 6 Rafał Mużyło 2010-02-28 19:50:32 UTC]

OK, have it your ways.

I wonder, however, if by "we do support them here in gentoo",
you're trying to say, that somebody tested if it actually works.

Or is that only "it does build" kind of support.

[Comment 7 Jory A. Pratt 2010-03-03 01:07:39 UTC]

I will review this myself just as soon as possible. If this is the case I will make the patch and send it upstream to remove --enable_ipv6 check from configure.

[Comment 8 Jory A. Pratt 2010-03-09 01:59:17 UTC]

ipv6 useflag drop'd in 4.8.4 which is in the tree now :) thanks for having us double check this.