Re: fix for 30094 breaks HURD compilation again

[Yavor Doganov]

Riccardo Mottola wrote:
> Well, I just check for GNU and GLIBC as
> http://glibc-bsd.alioth.debian.org/porting/PORTING
> suggests, which is a debian porting document!

Right, and I already corrected myself.  I confirm my initial patch
attached to the bug is not OK.

(For the record, I still retain my opinion that since the inception of
the kfreebsd* ports the majority of the packages have been doing this
change in order to build:

-#if defined __linux__ || defined __GNU__
+#if defined __linux__ || defined __GLIBC__

As I already mentioned, this implies GLIBC being defined by implicitly
including features.h, which is sometimes a false assumption.

Over the years, Debian has been changing many upstream code that
checks for generic (read: glibc-specific and sometimes not
glibc-specific, but kernel-independent) GNU features, that was relying
on __linux__ for one reason or another.  So, 5 or 10 years ago, the
correct fix was:

-#if defined __linux__
+#if defined __linux__ || defined __GNU__

...because GNU/Hurd was not in the picture, just like GNU/kFreeBSD
isn't right now.
)

> Please check if it works on kFreeBSD.

My GNU/kFreeBSD machine has been down for a year at least (hardware
problem), so the only way for me to test is to upload to Debian which
will trigger a rebuild on ~15 architectures.

> If troubles remain we can try your more complex macro patch.

I don't see how my patch is "more complex".  Removing the current
complexity, threreby simplifying the code (even in the trivial sense
by deleting lines) does not seem like additional complexity to me.
Quite on the contrary.

However, I completely agree with you that these things are not
trivial, and such changes should not be taken lightly.  I only asked
for volunteers to test my approach on various platforms (don't forget
to run `autoreconf' after applying the patch); I don't insist the
patch to be committed, even in an experimental branch, if it existed.
If it ever is, it should be thouroughly tested and would probably be
suitable only for the next stable release, not earlier.

I didn't want to be impudent at all, sorry if it sounded that way.

Anyway, your tweak is right given the current state of affairs, so I
certainly don't mind.  Just adjust the condition everywhere where
needed; that's all I ask.

(But have in mind that I'll nag you again when there is a
`kopensolaris' port :-))