"John P. Burkett" <burk...@uri.edu> posted 49fdcd09.7070...@uri.edu, excerpted below, on Sun, 03 May 2009 12:57:45 -0400:
> Thanks, Duncan. Yesterday I did eix-sync shortly before emerge --search > portage. To see if a new portage version became available overnight, I > just now did eix-sync and emerge --search portage. The results are the > same; the latest version available version of sys-apps/portage is still > listed as 2.1.6.11, which is the version I have installed. > > I attempted to manually download the source file, and place it in > distfiles, and then run emerge. Specifically, I downloaded > texlive-module-collection-latexextra-2008.tar.lzma from > http://ftp.ussg.iu.edu/linux/gentoo/distfiles/?C=N%3BO=D and placed the > file in /usr/portage/distfiles. Then I did "emerge texlive-latexextra". > The response stated with >>>> Verifying ebuild manifests >>>> Emerging (1 of 1) dev-texlive/texlive-latexextra-2008-r1 > [Errno 7] Argument list too long: > /bin/bash -c touch "/usr/portage/distfiles/.__portage_test_write__" > 2>/dev/null ; rval=$? ; rm -f > "/usr/portage/distfiles/.__portage_test_write__" ; exit $rval That's clearly a portage bug (even if we didn't already know it based on the bug you mentioned and the new versions that are /supposed/ to be out), as that argument list isn't even that long at all. So one way or another, we gotta get around that bug. One thing I noticed is that it's an lzma archive, which isn't all that common yet. You've verified that you can decompress that source archive manually, right? Meanwhile, on portage upgrade side... After a fresh sync to ensure I'm updated here, epkginfo portage shows the following: Keywords: portage-2.1.4.5: Keywords: portage-2.1.6.4: Keywords: portage-2.1.6.7: Keywords: portage-2.1.6.11: alpha amd64 arm hppa ia64 m68k ppc ppc64 s390 sh sparc x86 Keywords: portage-2.1.6.12: Keywords: portage-2.2_rc28: Keywords: portage-2.2_rc31: Keywords: portage-2.2_rc32: ~alpha ~amd64 ~arm ~hppa ~ia64 ~m68k ~mips ~ppc ~ppc64 ~s390 ~sh ~sparc ~sparc-fbsd ~x86 ~x86-fbsd So 2.1.6.12 is indeed in-tree, but no arch has keyworded it stable yet. I don't see any masking and checking the ebuild itself, I see it's keyworded ~arch. As it happens, I'm on ~arch and am running the 2.2-rc series. I had upgraded to rc32 before my first reply, and as has become my habit as a good admin, I checked the changelog before I upgraded. I thus noted mention of the fix for "bug #262647 ('Argument list too long' triggered by long SRC_URI)". So... I don't know why it hasn't been stable-keyworded, except that archs probably haven't gotten to it yet, but you might wish to consider adding: ~sys-apps/portage-2.1.6.12 ... to your package.keywords file or directory. (See the portage (5) manpage for details if you aren't familiar with package.keywords.) That will accept the unstable ~arch for just that one version of portage, including any revisions, should there be any necessary, thus allowing you to merge that version, which should have your bugfix. =:^) Since you're a stable user, presumably because you don't normally want the potential hassle of testing ~arch, I'd not recommend you just put portage in there (without a version), as there are indeed occasional issues ~arch users have to deal with, and when they're in the package manager... OTOH, ~arch users like me enjoy the challenge of being guinea pigs for stable users as well as getting newer packages even if it means occasional problems to deal with, and there's a place for both types of users. There are other alternatives too. Did you try using the --fetchonly option? The bug mentions that worked for some people. There's some additional discussion on why it happens -- are you using an old kernel (<2.6.23)? They had shorter max commandline lengths. Thus, upgrading your kernel is presumably another alternative. You can also try the specific patch on top of your current portage version. Of course, in this case it looks like you'd need SVN as all Zac mentions in the bug is the SVN revision number, but it's a common enough technique for folks wanting to keep tested and otherwise known to work versions. -- Duncan - List replies preferred. No HTML msgs. "Every nonfree program has a lord, a master -- and if you use the program, he is your master." Richard Stallman
