Created attachment 303849 [details] build log On my hardened ~amd64 machine with gcc-4.6.2 the test phase of boost fails while postprocessing the build log the error message is as follows. I checked and the file in question is on disk. I will now rerun with -j 1. **passed** ../bin.v2/libs/type_traits/test/has_operator_new_test.test/gcc-4.6/debug/pch-off/has_operator_new_test.test ...failed updating 208 targets... ...skipped 365 targets... ...updated 37585 targets... /var/tmp/portage/dev-libs/boost-1.48.0-r1/temp/environment: line 5774: tools/regression/build/bin/gcc-4.6/gentoorelease/pch-off/process_jam_log: No such file or directory * ERROR: dev-libs/boost-1.48.0-r1 failed (test phase): * Postprocessing the build log failed
Created attachment 303851 [details] emerge --info
Sorry Jer, but I think you are wrong in what causes this failure. If I "bzgrep utility /usr/share/doc/boost-1.47.0-r1/status/regress.log.bz2" I will get an awful lot of "../boost/tr1/detail/config_all.hpp:158:41: fatal error: ../4.6.2/utility: No such file or directory". But it is merged non the less. So this failure (which I also have) actually seems to be with something else.
Created attachment 304643 [details, diff] patch against ebuild to fix thinkos The src_test() part of the ebuild contains two thinkos: 1. The ebuild calls process_jam_log and compiler_status as if current directory is "${S}", not "${S}/status" which it really is. On the Other hand everything else is seems to be processed as if pwd really should be "${S}/status" so this seems like a thinko. 2. compiler_status wants the builddir as argument. However for some reason not giving the full path but only ".." makes compiler_status seek for something living in "../bin.v2/${S}" (the stdout output tells you otherwise but strace does not lie), so you really have to provide it with a full path... This patch makes those two commandos reference with ${S} (i.e. full paths) instead of trying to hit the relative paths. Makes the ebuild return from src_test() for me.
Could someone please change the summary, as the current one is totally not relevant!
I changed the summary according to what I understand of the current situation. I will probably try your patch later this evening or tomorrow.
The patch worked for me.
Thanks for the patch. Arfrever got to me first in IRC. + 12 Apr 2012; Mike Gilbert <floppym@gentoo.org> boost-1.48.0-r1.ebuild, + boost-1.49.0.ebuild: + Delete call to ebeep() (bug #409309). Fix paths to commands in src_test() (bug + #406467). Backported from Progress Overlay. Patch by Arfrever. +
*** Bug 407745 has been marked as a duplicate of this bug. ***
