The amount of bugs concerning ebuilds that ignore LDFLAGS suggests
that this would be a good idea, b/c it seems a many maintainers are
completely unaware that their ebuilds do not respect LDFLAGS - so I
guess this needs more visibility.

P.S. If you wonder why this flag then
check /usr/lib/portage/bin/misc-functions.sh

Cheers,

DN.

On Sun, 21 Mar 2010 00:46:38 +0100
Doktor Notor <notordoktor@gmail.com> wrote:

> The amount of bugs concerning ebuilds that ignore LDFLAGS suggests
> that this would be a good idea, b/c it seems a many maintainers are
> completely unaware that their ebuilds do not respect LDFLAGS - so I
> guess this needs more visibility.


does it work on non glibc/non linux systems?


Alexis.

On Saturday 20 March 2010 19:46:38 Doktor Notor wrote:
> The amount of bugs concerning ebuilds that ignore LDFLAGS suggests
> that this would be a good idea, b/c it seems a many maintainers are
> completely unaware that their ebuilds do not respect LDFLAGS - so I
> guess this needs more visibility.

remind me again why this matters ? binutils has been defaulting to hash-
style=both for quite a while now.
-mike

On Sun, 21 Mar 2010 00:52:36 -0400
Mike Frysinger <vapier@gentoo.org> wrote:

> remind me again why this matters ? binutils has been defaulting to
> hash- style=both for quite a while now.
> -mike

Well, I already tried in my P.S. but let me try again: I'm not
suggesting that devs should have more optimized binaries
- it matters b/c of this in misc-functions.sh:

if [[ "${LDFLAGS}" == *--hash-style=gnu* ]] && [[ "${PN}" != *-bin ]] ;
then;
<do some magic here to detect ignored LDFLAGS>
fi

Good candidate to test is net-fs/mount-cifs (tiny, fast)

You *only* get the following QA notice w/ LDFLAGS="-Wl,--hash-style=gnu"
set:

* QA Notice: Files built without respecting LDFLAGS have been detected
* Please include the following list of files in your report:
* /usr/bin/mount.cifs
* /usr/bin/umount.cifs

If you have another way to detect ignored LDFLAGS in portage, then you
don't need to add anything to devs profiles.

Hope that my proposal is more clear now.

Cheers,

DN

On Sunday 21 March 2010 02:15:17 Doktor Notor wrote:
> On Sun, 21 Mar 2010 00:52:36 -0400 Mike Frysinger wrote:
> > remind me again why this matters ? binutils has been defaulting to
> > hash- style=both for quite a while now.
>
> Well, I already tried in my P.S. but let me try again: I'm not
> suggesting that devs should have more optimized binaries
> - it matters b/c of this in misc-functions.sh:
>
> if [[ "${LDFLAGS}" == *--hash-style=gnu* ]] && [[ "${PN}" != *-bin ]] ;
> then;
> <do some magic here to detect ignored LDFLAGS>
> fi
>
> Good candidate to test is net-fs/mount-cifs (tiny, fast)
>
> You *only* get the following QA notice w/ LDFLAGS="-Wl,--hash-style=gnu"
> set:

forcing gnu-only hashes breaks some systems, and is a relatively "newish"
flag, so it isnt a candidate for global enabling.
-mike

On Sun, 21 Mar 2010 13:04:01 -0400
Mike Frysinger <vapier@gentoo.org> wrote:

> forcing gnu-only hashes breaks some systems, and is a relatively
> "newish" flag, so it isnt a candidate for global enabling.
> -mike

Well... am I really so difficult to understand?

I do *NOT* want it globally enabled... I want it enabled in
profiles/targets/developer/make.defaults *only*

Reason? Spit out the QA warnings about ignored LDFLAGS on maintainers.
Nothing else. Like there's already LD_AS_NEEDED="1" for detecting broken
--as-needed stuff. Is it finally clear now?

Cheers,

DN

On Sunday 21 March 2010 13:13:12 Doktor Notor wrote:
> On Sun, 21 Mar 2010 13:04:01 -0400 Mike Frysinger wrote:
> > forcing gnu-only hashes breaks some systems, and is a relatively
> > "newish" flag, so it isnt a candidate for global enabling.
> > -mike
>
> Well... am I really so difficult to understand?
>
> I do *NOT* want it globally enabled... I want it enabled in
> profiles/targets/developer/make.defaults *only*

so you suggest breaking some developers systems ? the relevance of my
statement still doesnt change.
-mike

On 21/03/10 18:13, Doktor Notor wrote:
> On Sun, 21 Mar 2010 13:04:01 -0400
> Mike Frysinger <vapier@gentoo.org> wrote:
>
>> forcing gnu-only hashes breaks some systems, and is a relatively
>> "newish" flag, so it isnt a candidate for global enabling.
>> -mike
>
> Well... am I really so difficult to understand?
>
> I do *NOT* want it globally enabled... I want it enabled in
> profiles/targets/developer/make.defaults *only*
>
> Reason? Spit out the QA warnings about ignored LDFLAGS on maintainers.
> Nothing else. Like there's already LD_AS_NEEDED="1" for detecting broken
> --as-needed stuff. Is it finally clear now?
>
> Cheers,
>
> DN


I am probably the only one, but I really don't like the way you are
talking! Please change that, otherwise I feel to do something against it!

Having good ideas or pointing out correct things, doesn't give anybody
the right to talk and act in the way you are doing! Not only here but
also in your comments in b.g.o.

Cheers, justin

On 03/21/2010 07:04 PM, Mike Frysinger wrote:
>> You *only* get the following QA notice w/ LDFLAGS="-Wl,--hash-style=gnu"
>> set:
>
> forcing gnu-only hashes breaks some systems, and is a relatively "newish"
> flag, so it isnt a candidate for global enabling.
> -mike

The developer profile target is being used only on linux profiles
(quickly grepped profiles) so I'm curious if it would break on one of
them...

On Sunday 21 March 2010 13:29:50 Samuli Suominen wrote:
> On 03/21/2010 07:04 PM, Mike Frysinger wrote:
> >> You *only* get the following QA notice w/ LDFLAGS="-Wl,--hash-style=gnu"
> >
> >> set:
> > forcing gnu-only hashes breaks some systems, and is a relatively "newish"
> > flag, so it isnt a candidate for global enabling.
>
> The developer profile target is being used only on linux profiles
> (quickly grepped profiles) so I'm curious if it would break on one of
> them...

it'll break all mips systems, it'll break on non-glibc systems, and it'll
break with older glibc/binutils systems.
-mike