Summary: | dev-lang/R-2.15.1 fails to install | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Patrick Lauer <patrick> |
Component: | New packages | Assignee: | Gentoo Science Mathematics related packages <sci-mathematics> |
Status: | RESOLVED FIXED | ||
Severity: | normal | CC: | ap, calchan, jlec, kripton |
Priority: | Normal | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Attachments: |
Log of failed build #1: R-2.15.2 with MAKEOPTS="-j9"
Log of failed build #2: R-2.15.2 with MAKEOPTS="-j9" Log of failed build with 'doc' use flag and MAKEOPTS="-j4" |
Description
Patrick Lauer
![]() Can you please retry with -j1? I'm really not sure it could be the issue here but you never know. There is a parallel compilation bug that neither us nor upstream have been able to fully identify. We thought we had nailed it at some point but it now appears that we haven't. If it has nothing to do that, any idea to give us a lead? I have installed R on many machines and apart from the odd and very rare parallel compilation issue I have never had any problem. Denis. I hit this yesterday with R-2.15.2. Then I tried R-2.15.1 which installed fine and R-2.15.2 again with MAKEOPTS="-j1" which worked then. Seems to be a race condition :( Same here with dev-lang/R-2.15.2 and hardened Gentoo amd64. With MAKEOPTS="-j1" R builds and instals fine and without it fails with the error mentioned above. I wonder what we should do with this. We have fixed parallel compilation issues in the past and we kind of suspected there was more. Our original reasonning was that the parallel compilation bug which we couldn't isolate was pretty rare so we'd leave parallel compilation in. Most, if not all, of the time re-emerging R without changing anything just worked. I can honestly not reproduce your issue here but I do believe you're having trouble with parallel compilation. We now probably need to investigate that more. One thing is that upstream is not very cooperative with such issues. If we fail to fix it we may have to disable parallel compilation entirely. Jannis, Coacher, do you systematically get the same error without MAKEOPTS="-j1"? In which case I'd be interested in looking at your build logs at least. Thanks, Denis. (In reply to comment #4) > Most, if not all, of the time re-emerging R without changing anything just > worked. > > Jannis, Coacher, do you systematically get the same error without > MAKEOPTS="-j1"? In which case I'd be interested in looking at your build > logs at least. Denis, the issue is gone itself. As you said, I've tried re-emerging and now everything builds and installs fine both with and without "-j1". (In reply to comment #5) > (In reply to comment #4) > > Most, if not all, of the time re-emerging R without changing anything just > > worked. > > > > Jannis, Coacher, do you systematically get the same error without > > MAKEOPTS="-j1"? In which case I'd be interested in looking at your build > > logs at least. > > Denis, the issue is gone itself. As you said, I've tried re-emerging and now > everything builds and installs fine both with and without "-j1". Not so here. Needed to rebuild R because icu-1.50 has been masked again and R failed again with MAKEOPTS="-j7" ... (In reply to comment #5) > Denis, the issue is gone itself. As you said, I've tried re-emerging and now > everything builds and installs fine both with and without "-j1". I don't think the issue is gone. What I think happened is you didn't hit it, because this parallel compilation issue is even more radom than usual. Which is why it has been so hard to fix. (In reply to comment #6) > Not so here. Needed to rebuild R because icu-1.50 has been masked again and > R failed again with MAKEOPTS="-j7" ... If you have time and interest in fixing this, could you please send us a build log? As I wrote above I can't promise I'll be able to fix it but I'll try. Thanks, Denis. Same problem here with the parallel build of R-2.15.2. It failed two times with MAKEOPTS="-j9" and built successfully afterwards with MAKEOPTS="-j1". I will attach the build logs... Created attachment 329366 [details]
Log of failed build #1: R-2.15.2 with MAKEOPTS="-j9"
Created attachment 329368 [details]
Log of failed build #2: R-2.15.2 with MAKEOPTS="-j9"
fellow gentooers, make sure the failure you get is in the install phase as shown in the Description (not reproducible for me), and not a duplicate of bug #395403. if you see in your build log a failure at compile time trying to compile cholmod, it is very likely you are hitting bug #395403. comment #9 and comment #10 are such dups. I think I ran into the same bug today. "cp: cannot create regular file '/var/tmp/portage/dev-lang/R-2.15.1/image//usr/include/Rmath.h': No such file or directory" during the 'make install' phase. For me it appears to be related to the 'doc' use flag. I've emerged dev-lang/R-2.15.2-r2 successfully twice, both without the 'doc' use flag. The only difference between these two merges was the 'cairo' use flag. Both these merges were with MAKEOPTS="-j4". However trying to emerge with the 'doc' use flag enabled failed twice with the above error when using MAKEOPTS="-j4". I will attach a build log of one of these failures. When I changed to MAKEOPTS="-j1" there was no error. Created attachment 340012 [details]
Log of failed build with 'doc' use flag and MAKEOPTS="-j4"
(bzip2'd since it was over the allowed filesize limit)
(In reply to comment #12) > However trying to emerge with the 'doc' use flag enabled failed twice with > the above error when using MAKEOPTS="-j4". I will attach a build log of one > of these failures. I tried to reproduce this behaviour on a different installation, but there the install went fine with the same use flags (so including 'doc') and MAKEOPTS="-j4". This pc is a little bit slower than the pc on which I first encountered the problem. (In reply to comment #14) > (In reply to comment #12) > > However trying to emerge with the 'doc' use flag enabled failed twice with > > the above error when using MAKEOPTS="-j4". I will attach a build log of one > > of these failures. > I tried to reproduce this behaviour on a different installation, but there > the install went fine with the same use flags (so including 'doc') and > MAKEOPTS="-j4". > > This pc is a little bit slower than the pc on which I first encountered the > problem. And to add another data point: Also on a third installation (which is a bit slower than the first) I could not reproduce this behaviour. Again R installed fine with the same use flags (including 'doc') and MAKEOPTS="-j4". + 15 Jan 2014; Justin Lecher <jlec@gentoo.org> -R-2.10.1.ebuild, + -R-2.15.3.ebuild: + Drop old; fixes #478572 (EAPI <4 should be removed), #476010 & #476008 + (Respect AR and LDFLAGS), #438544 (Installation Problems with version 2), + #425708 (stack smashing) + |