Issue 116423 - [gbuild] ld hash-style both
Summary: [gbuild] ld hash-style both
Status: CLOSED FIXED
Alias: None
Product: Build Tools
Classification: Code
Component: GNU make (show other issues)
Version: DEV300m96
Hardware: All All
: P3 Normal (vote)
Target Milestone: 3.4.1
Assignee: pavel
QA Contact: issues@tools
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2011-02-24 14:15 UTC by hans_werner67
Modified: 2017-05-20 10:31 UTC (History)
4 users (show)

See Also:
Issue Type: DEFECT
Latest Confirmation in: ---
Developer Difficulty: ---

Attachments
Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this issue.
Description pavel 2011-01-12 11:13:51 UTC 
configure prints:

checking for --hash-style=both linker support ... not found

unxlng.mk has:

.IF "$(HAVE_LD_HASH_STYLE)"  == "TRUE"
LINKFLAGS += -Wl,--hash-style=both
.ELSE
LINKFLAGS += -Wl,-zdynsort
.ENDIF

but solenv/gbuild/platform/linux.mk has:

	-Wl,--hash-style=both \

Please make this conditional as well.
Comment 1 mst.ooo 2011-01-13 10:22:46 UTC 
another one for b_michaelsen....

i wonder: since when is this --hash-style=both support in binutils?
is it missing only on ancient distros,
or did it fail for you (pjanik) on something that isn't so old?
Comment 2 pavel 2011-01-13 12:08:58 UTC 
My build machine is Mac OS X Tiger.
Comment 3 bjoern.michaelsen 2011-01-13 12:24:40 UTC 
accepting (Status -> STARTED)
adding issue to cws gnumake3
Comment 4 Stephan Bergmann 2011-01-13 13:18:32 UTC 
@pjanik: "solenv/gbuild/platform/linux.mk has" vs. "build machine is Mac OS X
Tiger"?!?  ;)
Comment 5 pavel 2011-01-13 13:42:22 UTC 
ok, ok, I intermixed two issues :-)

This issue is on some old SUSE chroot.I think it is 10.0.
Comment 6 bjoern.michaelsen 2011-01-13 16:07:58 UTC 
Unconditionally falling back on some setting that is rather new (dynsort) and
still undocumented in the ld manpage, when configure tells us that ld does not
even support --hash-style=both (and that test looks sensible in configure.in) is
... weird. Not using the feature when we have --hash-style=both support is also
... weird (unless it it enabled by default now, but still).

@pjanik: Do have any insight on this? Or can mmeeks provide some as he seems to
have implemented dynsort for ld?
Comment 7 mst.ooo 2011-01-13 16:15:12 UTC 
@mmeeks:
do you happen to know whether this fallback
from --hash-style=both to -zdynsort makes sense?
Comment 8 mmeeks 2011-01-13 17:59:45 UTC 
-zdynsort was rather useful - but only before --hash-style=both arrived. I
believe its deployment is rather tiny now (I hope so anyway), so - I would fall
back to nothing. Though, you really want -zcombreloc (though I think we had that
already).

Hope that helps :-)
Comment 9 bjoern.michaelsen 2011-01-14 00:35:17 UTC 
Thanks for the hints.
Ok, so we will simply fall back to nothing in this case.
As for -zcombreloc, my google foo suggests it is enabled by default at least
since binutils 2.15.
Comment 10 bjoern.michaelsen 2011-02-02 09:25:19 UTC 
fixed in gnumake3 (hash-style-both when available, fallback to nothing
otherwise). -zcombreloc is always set, although it is on by default on a modern
toolchain.
Comment 11 bjoern.michaelsen 2011-02-02 09:26:15 UTC 
@pjanik: please verify
Comment 12 Oliver-Rainer Wittmann 2012-06-13 12:31:48 UTC 
getting rid of value "enhancement" for field "severity".
For enhancement the field "issue type" shall be used.