There's a sandbox violation during installation of silc-toolkit 0.9.12-r3 because it tries to create the directory /usr/tutorial directly. Here's the output: mkdir -p /var/tmp/portage/silc-toolkit-0.9.12-r3/image//usr/share/doc/silc-toolkit-0.9.12-r3/toolkit/ /bin/install -c -m 644 ./doc/toolkit/* /var/tmp/portage/silc-toolkit-0.9.12-r3/image//usr/share/doc/silc-toolkit-0.9.12-r3/toolkit /bin/install -c -m 644 ./lib/doc/*.gif /var/tmp/portage/silc-toolkit-0.9.12-r3/image//usr/share/doc/silc-toolkit-0.9.12-r3/toolkit cp -R ./tutorial /usr ACCESS DENIED mkdir: /usr/tutorial cp: cannot create directory `/usr/tutorial': Permission denied make[3]: [toolkit-install] Error 1 (ignored) [... I did cut here ...] >>> Completed installing into /var/tmp/portage/silc-toolkit-0.9.12-r3/image/ --------------------------- ACCESS VIOLATION SUMMARY --------------------------- LOG FILE = "/tmp/sandbox-net-im_-_silc-toolkit-0.9.12-r3-11389.log" mkdir: /usr/tutorial --------------------------------------------------------------------------------
This works fine for me now. Looks like install locations were recently fixed. Can you sync and try again?
I just tried, after syncing, still does not work for me: FEATURES="buildpkg ccache digest sandbox userpriv usersandbox" I may add I'm running an amd64 machine.
CC'ing ticho as this is his baby. Ticho feel free to re-assign to yourself.
Jocelyn, could you please post your entire log file for this package from /var/log/portage/ ?
Just retried the merge. Here are infos given by emerge --pretend: [ebuild U ] net-im/silc-toolkit-0.9.12-r3 [0.9.12-r2] -debug -debug +ipv6 0 kB I also got +doc in my USE (may be related to this, as it concerns /usr/doc) There no more info in the sandbox log file, sorry: # cat /tmp/sandbox-net-im_-_silc-toolkit-0.9.12-r3-654.log mkdir: /usr/tutorial #
No, I meant a complete log from emerge process, portage stores them in /var/log/portage by default. For example, log from my last emerge of silc-toolkit is /var/log/portage/3425-silc-toolkit-0.9.12-r3.log. Also, please check md5sum of the ebuild file, so we can strike out the possibility that you are using incorrect ebuild (maybe mirror out of date?). I have changed the -r3 ebuild few times within short period of time, and I have played mainly with installation paths, so chances are the mirror picked up not-so-good ebuild. My (working) ebuild has md5sum like this: 00af6731f4f63adffd40af191be564f6 silc-toolkit-0.9.12-r3.ebuild You can use md5sum command for this (in case you didn't know).
I already posted the output of emerge concerning the problem in my first message. The content of the log file is exactly the same and there no other information related to the problem in the whole output. The only thing that may be useful is the end of the make output: (comes just after the place where I noted 'I did cut here') mkdir -p /var/tmp/portage/silc-toolkit-0.9.12-r3/image//usr/share/doc/silc-toolk it-0.9.12-r3/examples/ /bin/install -c -m 644 ./doc/examples/README /var/tmp/portage/silc-toolkit-0.9.1 2-r3/image//usr/share/doc/silc-toolkit-0.9.12-r3/examples/ /bin/install -c -m 644 ./doc/examples/silc* /var/tmp/portage/silc-toolkit-0.9.12 -r3/image//usr/share/doc/silc-toolkit-0.9.12-r3/examples/ make[3]: Leaving directory `/var/tmp/portage/silc-toolkit-0.9.12-r3/work/silc-to olkit-0.9.12' make[2]: Leaving directory `/var/tmp/portage/silc-toolkit-0.9.12-r3/work/silc-to olkit-0.9.12' make[1]: Leaving directory `/var/tmp/portage/silc-toolkit-0.9.12-r3/work/silc-to olkit-0.9.12' What I can see is the md5sum of the file I got is not the same as yours. I doubt the sync server is not up-to-date, because I sync every night and update my packages daily. If the server stopped updating its base, I think I would not have new/updated packages every day, but I have. As sync servers are rotating, it would be strange that no server is up-to-date... I use: SYNC=rsync://rsync.europe.gentoo.org/gentoo-portage The sync for the last five days were done using: 09-16: rsync://80.190.246.242/gentoo-portage 09-15: rsync://160.45.117.66/gentoo-portage 09-14: rsync://130.230.54.100/gentoo-portage 09-13: rsync://134.147.32.57/gentoo-portage 09-12: rsync://134.147.32.57/gentoo-portage Here are the md5s I got: # md5sum /usr/portage/net-im/silc-toolkit/*.ebuild 9a5e70dcc2b8cf19a19eb000d52e022d /usr/portage/net-im/silc-toolkit/silc-toolkit-0.9.12-r2.ebuild 3abc19103b2f36417fba2a882dcc3361 /usr/portage/net-im/silc-toolkit/silc-toolkit-0.9.12-r3.ebuild # Hope this can help.
The md5sums match, I forgot that I added a single comment line to the my local copy of the ebuild, sorry about that. That said, I really need the log I requested, as I need to check if certain variables were set up correctly near the beginning. Could you _please_ attach the corresponding log? Thanks a lot.
Created attachment 39719 [details] emerge failure log file
Created attachment 39722 [details] testing ebuild I think I know where the problem lies. first sed command in ebuild's src_unpack() function doesn't do anything on your system. Could you try this modified ebuild to see if it fixes the problem? Also, I'm interested in output of `sed --version` on your system.
OK, I found it. It's not a sed bug. When I do ebuild .... unpack, Makefile.in is correctly patched. But, for a reason I ignore, the build process recreates it from Makefile.am. So I duplicated the sed command to patch Makefile.am, and the problem disapeared.
Brilliant, thanks for the fix! Updated -r3 ebuild is now in CVS portage, should be hitting mirrors in few hours. To speed things up, I'm attaching the ebuild I committed here, so you can test right away. So, see you on SILCNet :)
Created attachment 39751 [details] fixed ebuild
I tested exactly the same patch (I just got the comment in the diff), I re-tested to be sure: I can confirm it solves the problem. Thanks for all !
