Summary: | dev-qt/qtgui-5.5.1 fails to build - QT_MGE_PGE_PATTERN: own own placehol%{if%{ifQIODe::re::read (QFi"../"../3 Cannot process input: '../3rdpardpa'. Stop. | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Sernin van de Krol <serninpc> |
Component: | [OLD] Library | Assignee: | Qt Bug Alias <qt> |
Status: | RESOLVED DUPLICATE | ||
Severity: | normal | CC: | josef64 |
Priority: | Normal | ||
Version: | unspecified | ||
Hardware: | x86 | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Attachments: | build.log |
Description
Sernin van de Krol
2015-12-16 15:41:09 UTC
Created attachment 419370 [details]
build.log
> KiB Mem: 247500 total, 38064 free
did you run out of memory?
I don't think I ran out of memory. There is a 1 GB swap partition and nothing in dmesg or syslog to indicate out of memory. Only things running were syslog-ng, cronie and sshd. Running that last command from a shell gives the same output. Hm, would this builds with latest stable gcc-4.9.3 ? Same error with the stable gcc 4.9. Remerging qtdbus (the package with qdbusxml2cpp) with the new gcc didn't help either. Also, I searched for QT_MGE_PGE_PATTERN and came across at least one case of faulty memory. To be sure I am now running memtest86+ on the machine, it has currently completed 6 passes without errors. I did some more research and got some new, weird, more usable results: I recompiled sys-apps/dbus and dev-qt/qtcore with gcc 4.9. Those were the first candidates, because glibc should probably give issues much earlier if something is wrong. Then I recompiled dev-qt/qtdbus, and tried to run it in the leftover directory from the last compile attempted by portage. qdbusxml2cpp started printing the following message in an infinite loop: Incompatible processor. This Qt build requires the following features: sse2 Weirdest thing ever, cause my pentium 3 has pentium3 as march. I know -march=native is the preferred way, but is unusable if you want to do stuff with distcc. The instructions I followed during setup are those found on the Gentoo Wiki: https://wiki.gentoo.org/wiki/Safe_CFLAGS#Find_CPU-specific_options I don't know how any sse2 instructions would have gotten in with my current CFLAGS, but I am tempted to think something in de Qt build system is enabling it, because it seems to know about using them. Also there is this line in the build.log supporting my claim: i686-pc-linux-gnu-g++ -c -O2 -march=pentium3 -mno-sahf -msse2 -mfpmath=sse And my CFLAGS: CFLAGS="-O2 -march=pentium3 -mno-sahf" Looking at the previous lines of the build.log reveals Qt assuming a lot of things about the processor. (I cut some stuff for clarity.) Build options: Configuration .......... *snip* avx avx2 *snip* sse2 sse3 sse4_1 sse4_2 ssse3 *snip* Which is strange, because above that it clearly states: Build type: linux-g++ (i386, CPU features: mmx sse) Thanks for the analysis. This is a known bug. *** This bug has been marked as a duplicate of bug 552942 *** |