I'm running into some very strange behavior trying to build plplot which uses cmake.
Scratch build here: http://koji.fedoraproject.org/koji/taskinfo?taskID=765779
i386 build log (on x86-7 - real host?): http://koji.fedoraproject.org/koji/getfile?taskID=765782&name=build.log
x86_64 build log (on xenbuilder2 - xen host): http://koji.fedoraproject.org/koji/getfile?taskID=765781&name=build.log
+ /usr/bin/cmake -DCMAKE_INSTALL_PREFIX:PATH=/usr -DCMAKE_INSTALL_LIBDIR:PATH=/usr/lib64 -DINCLUDE_INSTALL_DIR:PATH=/usr/include -DLIB_INSTALL_DIR:PATH=/usr/lib64 -DSYSCONF_INSTALL_DIR:PATH=/etc -DSHARE_INSTALL_PREFIX:PATH=/usr/share -DLIB_SUFFIX=64 -DBUILD_SHARED_LIBS:BOOL=ON .. -DENABLE_ada:BOOL=ON -DENABLE_d:BOOL=ON -DENABLE_ocaml:BOOL=ON -DENABLE_pdl:BOOL=ON -DHAVE_PTHREAD:BOOL=ON -DPL_FREETYPE_FONT_PATH:PATH=/usr/share/fonts/freefont -DPLD_aqt:BOOL=ON -DPLD_conex:BOOL=ON -DPLD_dg300:BOOL=ON -DPLD_imp:BOOL=ON -DPLD_linuxvga:BOOL=ON -DPLD_ljii:BOOL=ON -DPLD_ljiip:BOOL=ON -DPLD_mskermit:BOOL=ON -DPLD_ntk:BOOL=ON -DPLD_pstex:BOOL=ON -DPLD_svg:BOOL=ON -DPLD_tek4010:BOOL=ON -DPLD_tek4010f:BOOL=ON -DPLD_tek4107:BOOL=ON -DPLD_tek4107f:BOOL=ON -DPLD_versaterm:BOOL=ON -DPLD_vlt:BOOL=ON -DPLD_xterm:BOOL=ON -DPLD_wxwidgets:BOOL=ON -DBUILD_DOC:BOOL=ON -DBUILD_TEST:BOOL=ON Linux 2.6.18-92.1.6.el5xen
and it hangs here. I looks like cmake is running "uname" through popen but the output goes to the output and cmake is stuck reading a bad file descriptor.
I can reproduce here on my F8 x86_64 xen builder, but it hangs on the i386 build not the x86_64. strace of the hung cmake shows:
read(4286667132, 0x55577000, 8192) = -1 EBADF (Bad file descriptor)
over and over.
Anyone else seeing anything like this?
Sorry I don't have anything more concrete but I'm out for a long weekend and thought I'd get this out there...
Orion Poplawski <orion <at> cora.nwra.com> writes:
I'm running into some very strange behavior trying to build plplot which uses cmake.
We've also seen this with KDE 4 builds (perhaps not surprisingly, as KDE 4 also uses cmake). Just yesterday I had to cancel and resubmit a konq-plugins build for that reason, I wanted to report it, but you were quicker. :-)
IIRC, I have only seen this on F10 x86_64 builds.
Kevin Kofler
Orion Poplawski <orion <at> cora.nwra.com> writes:
I'm running into some very strange behavior trying to build plplot which uses cmake.
We've also seen this with KDE 4 builds (perhaps not surprisingly, as KDE 4 also uses cmake). Just yesterday I had to cancel and resubmit a konq-plugins build for that reason, I wanted to report it, but you were quicker. :-)
IIRC, I have only seen this on F10 x86_64 builds.
Any idea when it started? I built cmake 2.6.1 on the 5th...
- Orion
Kevin Kofler wrote:
Orion Poplawski <orion <at> cora.nwra.com> writes:
I'm running into some very strange behavior trying to build plplot which uses cmake.
We've also seen this with KDE 4 builds (perhaps not surprisingly, as KDE 4 also uses cmake). Just yesterday I had to cancel and resubmit a konq-plugins build for that reason, I wanted to report it, but you were quicker. :-)
IIRC, I have only seen this on F10 x86_64 builds.
Filed bug #459442 as I have a simple test case. Once everything is back up we can test again.
Orion Poplawski wrote:
Filed bug #459442 as I have a simple test case. Once everything is back up we can test again.
It appears that the pipe2 syscall on the x86_64 xen kernels is broken and that rawhide glibc has moved to using pipe2 from pipe in rawhide.
This seems like a blocker to me.
Orion Poplawski wrote:
Orion Poplawski wrote:
Filed bug #459442 as I have a simple test case. Once everything is back up we can test again.
It appears that the pipe2 syscall on the x86_64 xen kernels is broken and that rawhide glibc has moved to using pipe2 from pipe in rawhide.
This seems like a blocker to me.
Okay, looks like a known issue with the xen kernels. New test kernels are available, see the bug for more info. We should get a fixed version onto the Fedora xen builders very soon, and maybe disable them until they can be fixed?