On 04/19/2018 02:56 PM, Stephan Bergmann wrote:
> Not sure how -Wclass-memaccess is relevant for LTO?
It's much easier to end up with an UBSAN that eventually causes a SEGFAULT.
That's my experience learned from Firefox.
Anyway, had cleaned up LO master the other day to compile with upcoming GC
On 04/10/2018 09:32 AM, Tor Lillqvist wrote:
> The help message for the --enable-lto option says: "This is experimental work
> in progress that shouldn't be used unless you are working on it."
>
> Thus you have two choices: 1) Don't do that then, or 2) Debug and fix the
> problem.
>
> --tml
Hi
Following test-case is failing with --enable-lto:
[ 7806s] /bin/sh: line 1: 1652 Segmentation fault (core dumped) (
LD_LIBRARY_PATH=${LD_LIBRARY_PATH:+$LD_LIBRARY_PATH:}"$I/program:$I/pro
gram":$W/UnpackedTarball/cppunit/src/cppunit/.libs MALLOC_CHECK_=2
MALLOC_PERTURB_=153 $W/LinkTarget/Exe
solenv/gbuild/platform/com_GCC_defs.mk |2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
New commits:
commit d7cb910ad888401335552835314d689a49351924
Author: Martin Liška
Date: Wed Mar 26 13:22:05 2014 +0100
gbuild: enable LTO parallelism for GCC
Change-Id: Id0be70056645a
Sure,
there's correct patch.
Martin
On 03/26/2014 12:59 PM, Miklos Vajna wrote:
Hi Margin,
On Wed, Mar 26, 2014 at 11:43:24AM +0100, Martin Liška wrote:
diff --git a/solenv/gbuild/platform/com_GCC_defs.mk
b/solenv/gbuild/platform/com_GCC_defs.mk
index a4b18a3..93d3fe7 100644
Hello,
following patch increases parallelism for LTO. I tried to add
-flto=+jobserver, but didn't help.
Tested on trunk for x86_64 GNU/Linux.
Thank you,
Martin
diff --git a/solenv/gbuild/platform/com_GCC_defs.mk b/solenv/gbuild/platform/com_GCC_defs.mk
index a4b18a3..93d3fe7 100644
--- a/so
I built libreoffice with disabled LTO and parallel build was successfull :)
That means LTO optimization really corrupted some of libreoffice
libraries and as a result tests are failing?
Thank you,
Martin
On 3 June 2013 10:34, Michael Meeks wrote:
>
> On Mon, 2013-06-03 at 00:41 +0200,
t I've only done it once or twice.
> The multiple definition errors should all be fixed now.
>
> Thanks,
>
> Peter
>
> On Mon, Mar 25, 2013 at 5:25 AM, Jan Holesovsky wrote:
>> Hi Martin
>>
>> Martin Liška píše v Ne 24. 03. 2013 v 20:26 +0100:
>>
>&g
