The latest pkgbuild for gnofract4d fails to build with a segmentation fault on my machine. (AUR page has details.) This looks like the same issue I had with an earlier version (details on the same page). There is a note in createdocs.py which states
# making this a separate cmd (not part of setup.py) because importing
# gtk was causing "setup.py build" to crash - inexplicable...
# also the other python versions don't have gtk as a module,
# so were reporting errors.
My output suggests that it is failing because I'm compiling in an account which does not have access to DISPLAY. That seems an odd thing for a script which is meant to produce package documentation to require. Is this a known bug/feature? Is there any way around it (other than giving it access)?
I should explain that I compile all AUR packages in a relatively unprivileged account (which does not have access to sudo not being in wheel etc.). I don't log in as this user - I just su to the account and compile. (This user also manages most software in /usr/local except for TeX Live which has a dedicated user to manage it.) I do this to limit the damage which a malicious/inept pkgbuild could do i.e. such a recipe could do rm -rf ~/* but it wouldn't really matter because it would not be the home of a real user i.e. it isn't my home and doesn't contain my documents and data etc. Of course, I hope I would catch anything that blatant and since I have to install packages with root privileges, this doesn't eliminate risk, but it does do something to limit the damage which might be done via one possible route.
I don't altogether trust a pkgbuild or an install script which wants access to DISPLAY in order to compile the docs. Why would it need GUI access to create a man page? So I am reluctant to give it carte blanche...
Last edited by cfr (2013-07-24 00:37:44)