On 12/05/2016 23:13, Ken Moffat wrote: > On Thu, May 12, 2016 at 09:59:38PM +0200, Pierre Labastie wrote: >> >> I'll try those. Right now, I do not have (graphical) access to the machine >> where I have built qupzilla. From memory, it may happen that the first >> visited >> site may last longer than 5 sec, but after a few link followings, pages first >> load, then disappear with the message: "cannot load page, try to free close >> tabs to free memory, and try to reload". According to top, a lot of memory is >> free. When I hit the "reload" button, the page appears for a few seconds, and >> then dissappears again. >> >> I have no time to look at this tonight. I'll try to do something tomorrow, >> but >> I may have a lot of other things to do, so not sure. >> >> Pierre >> > > There is no rush, particularly when you are doing this as a favour. > > I can't see any text resembling that in either qtwebengine or > qupzilla, and grepping through the rest of qt hits too many files or > directories with spaces in their names. > > Is this 32-bit (small addressable memory space) ? > > If you have not tried running it from your preferred term, maybe > that might put some messages out. > > Hmm, I wrote that while doing the grepping. But then I tried > googling with different terms and eventually got to: > https://bodhi.fedoraproject.org/updates/FEDORA-2016-a62ff1916b > > quote: > > · The C++ files are now built with -fno-delete-null-pointer-checks > to work around a miscompilation by GCC 6 (#1313575) that had made > the package mostly useless. > > · The package now includes working debugging information. (Note: > Full debugging information is only included in the x86_64 builds. > The 32-bit builds (i686, armv7hl) include only the debugging > information required for backtraces (-g1) because the linker runs > out of address space trying to link libQtWebEngineCore with full > debugging information on 32-bit.) > > Their specfile is at > http://pkgs.fedoraproject.org/cgit/rpms/qt5-qtwebengine.git/plain/qt5-qtwebengine.spec > and their patches are at > http://pkgs.fedoraproject.org/cgit/rpms/qt5-qtwebengine.git/tree/ > > The spec has the following three seds which might perhaps be useful, > or might not be needed outside of fedora. > > # fix // in #include in content/renderer/gpu to avoid debugedit failure > sed -i -e 's!gpu//!gpu/!g' \ > > src/3rdparty/chromium/content/renderer/gpu/compositor_forwarding_message_filter.cc > # remove ./ from #line commands in ANGLE to avoid debugedit failure (?) > sed -i -e 's!\./!!g' \ > > src/3rdparty/chromium/third_party/angle/src/compiler/preprocessor/Tokenizer.cpp > \ > > src/3rdparty/chromium/third_party/angle/src/compiler/translator/glslang_lex.cpp > > %ifnarch x86_64 > # most arches run out of memory with full debuginfo, so use -g1 on > # non-x86_64 > sed -i -e 's/=-g$/=-g1/g' src/core/gyp_run.pro > %endif > > But the main item is > export CXXFLAGS="%{optflags} -fno-delete-null-pointer-checks" > %ifnarch x86_64 > # most arches run out of memory with full debuginfo, so use -g1 on > # non-x86_64 > export CXXFLAGS=`echo "$CXXFLAGS" | sed -e 's/ -g / -g1 /g'` > %endif > > i.e. just add -fno-delete-null-pointer-checks to CXXFLAGS and on > 32-bit, if CXX flags include -g change it to -gl. Sorry if I'm > telling you things you already knew, but I think you said you were > not familiar with fedora. > > ĸen >
Sorry for the delay in replying. Thanks a lot for this post. It's great you found the fedora bug report. This "-fno-delete-null-pointer-checks" addition is the same as that for OpenJDK. I understand it removes some too agressive optimization. It seems that only qtwebengine needs that, and not the whole Qt, so I wonder how to do it: First build Qt with -skip-qtwebengine, then run qmake for qt webengine (as in the fedora spec above), and make again? Or add a new page for qtwebengine? Anyway, for now, the easiest for me is to build qt with -skip-qtwebengine, then come back later to qtwebengine. I'll let you know how I'll exactly do, and how qupzilla works... Pierre -- http://lists.linuxfromscratch.org/listinfo/blfs-dev FAQ: http://www.linuxfromscratch.org/blfs/faq.html Unsubscribe: See the above information page
