I redirect outputs of emerge to a file. first steps of ebuild are quickly done. last message written to logfile is 'X Translate binary manpages to ca'. CPU consumption increases to 100%. over time also memory consumption increases. after 12 hours the amount of memory used by the build process reached 1.2 GB. I had to stop emerge using CTRL-C. after some investigations doing a private build (unpacked src; ran 'make' (which created new perl script 'Build'); ran ./Build), which succeeded. tried emerge again but w/o output redirection - succeeded. removed po4a and emerged again w/ output redirection: hung again. final emerge w/o output redirection succeeded again. I guess that issues 257741 and 219889 might have similar problems. Reproducible: Always Steps to Reproduce: 1. emerge po4a > po4a.log 2>&1 Actual Results: hangs with 100% CPU (maybe only until out-of-memory) Expected Results: ebuild should work too if output redirection is used installed version of perl is 5.8.8-r5 installed version of perl-Module-Build is 0.28.08
Created attachment 191127 [details] build.log Confirmed on amd64. Works without redirection, hangs with redirection.
It's the perl process that hangs, not emerge.
for me it hangs at X Translate binary manpages to pl without redirection. not sure if this problem is related. x86 here.
(In reply to comment #3) > for me it hangs at > > X Translate binary manpages to pl > > without redirection. not sure if this problem is related. x86 here. > Hi, assuming you are using paludis, this is the same problem, yeah. :) Best regards, Mihai
No solution yet? I have the same problem with paludis and x86 and wonder how I can install dpkg now...
Are there any know workarounds? I guess we're hitting lack-of-time atm mainly, me included. Stephan, do you have time to help out with further research?
Created attachment 203139 [details, diff] fix-io-capture-0.36.patch After applying the attached patch by extending PATCHES to read PATCHES=( "${FILESDIR}/0.36-3-debian.patches" "${FILESDIR}/fix-io-capture-${PV}.patch" ) It ceases to fail on me during build, and instead progresses nicely and passes the test suite. Why exactly It fails however, is still a bewilderment.
Created attachment 203140 [details] Modified ebuild
( nb. tests performed on perl5.10.1 , you're experience may differ, but I don't think anything reasonable in perl changes since 5.8.8 would translate from 12 hours to : app-text/po4a: Sat Sep 5 07:15:36 2009: 1 minute, 11 seconds )
Also to note, an alternative of system(@list) does *NOT* work either. open(-|) appears to be the only thing that works.
0.36.4 with the patch is in the tree. I am closing this bug and leave the tracker bug open until we find the real cause
Hi, is there a way to create a patch for 0.32? (The workaround for me as a paludis-user was to temporarily switch back to emerge) The 0.36.4 depends (indirectly) on perl5.10.1, which is not even in the tree, and I really don't need another overlay :) Thanks