391937 - Linux x86_64 startup-notification relocation error gcc-4.2.x

Status: RESOLVED FIXED

(Firefox Build System :: General, defect)

Description

[Walter Meinl]

User-Agent:       Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.9a8pre) Gecko/2007081215 Minefield/3.0a8pre
Build Identifier: Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.9a8pre) Gecko/2007081215 Minefield/3.0a8pre

/usr/lib/gcc/x86_64-pc-linux-gnu/4.2.0/../../../../x86_64-pc-linux-gnu/bin/ld: n
sWindow.o: relocation R_X86_64_PC32 against `sn_display_new' can not be used whe
n making a shared object; recompile with -fPIC
/usr/lib/gcc/x86_64-pc-linux-gnu/4.2.0/../../../../x86_64-pc-linux-gnu/bin/ld: f
inal link failed: Bad value

Reproducible: Always

Steps to Reproduce:
1. compile minefield with gcc > 4.2
2.
3.
Actual Results:  
relocation error

Expected Results:  
should compile

Comment 1

[Walter Meinl]

Attachment: add the sn headers to the system-wrappers

Comment 2

[Benjamin Smedberg]

Comment on attachment 276376 [details] [diff] [review]
add the sn headers to the system-wrappers

Please use #ifdef instead of ==

Comment 3

[Walter Meinl]

Attachment: use ifdef (all occasions)

In reply to comment #2)
> (From update of attachment 276376 [details] [diff] [review])
> Please use #ifdef instead of ==
> 
Ups, when I looked at config/system-headers I thought this was desired

benjamin, in this patch version I changed all '==' or '!=' against ifdef or ifndef, working with the sn headers and with native-lcms zlib png and jpeg (cairo not tested, but should work as well)
Since I'm German I want it standardized ;-)
When you think it's overdone I'll attach a patch only ifdef(ed) for startup-notification headers

Comment 4

[Walter Meinl]

Comment on attachment 276513 [details] [diff] [review]
use ifdef (all occasions)

asking for approval for the trunk. Should be very low risk - and a helper for checkin (if approval +) thanks

Comment 5

[gentoo_e17_powa]

I am under Gentoo ~amd64 and I have so build errors when emerging Firefox-2.0.0.6, Thunderbird 2.0.0.6 and Xulrunner 1.8.1.6, with gcc-4.2.0, glibc-2.6.1, binutils-2.17.50.0.18.
I have had to disable the "debug" USE flag to build this software.
See : https://bugs.gentoo.org/show_bug.cgi?id=188636

Comment 6

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 276513 [details] [diff] [review]
use ifdef (all occasions)

a=bzbarsky

Comment 7

[Boris Zbarsky [:bzbarsky]]

Checked in on trunk.  Thanks for the patch!

Comment 8

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

I think those changes to use ifdef will break things, given the Makefile.  If you want to use ifdef, you have to convert the code that passes through various AC_DEFINEs from configure.in as -Ds to not pass them at all.  So I think what landed will break a bunch of things.

Did anyone test that it didn't?

Comment 9

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

And note that to test this you need to remove the system-headers directory and clobber any directories that depend on the wrapping.

Comment 10

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Yeah, I confirmed locally that the patch broke those tests.  The changes to existing ifdefs should be backed out; the new stuff is fine.

Comment 11

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

I backed out the ifdef changes in existing code since they broke things (those variables were always defined, sometimes to empty string).

Comment 12

[Walter Meinl]

(In reply to comment #11)
> I backed out the ifdef changes in existing code since they broke things (those
> variables were always defined, sometimes to empty string).
> 

Thanks that you kept an eye on it, I enforced it too much