<farkuhar>
Today I was investigating FS#1933 and hit an error building contrib/farstream. It was not easy to disable the dependency on gtk-doc, but after wrestling with autotools I came up with this: http://sprunge.us/FD4mEg
<farkuhar>
After solving the farstream problem, I was able to build pidgin, but I couldn't reproduce the nss error reported in FS#1933.
<farkuhar>
Heh, the first sed in my revised Pkgfile wasn't really necessary. But in order to test pidgin for real, I'll probably have to do more than just launch it, and actually connect to a network.
farkuhar has left #crux-devel [#crux-devel]
farkuhar has joined #crux-devel
farkuhar has left #crux-devel [#crux-devel]
farkuhar has joined #crux-devel
<farkuhar>
I finally got pidgin to crash as described in FS#1933, by asking it to use SSL when connecting to the server. This current connection is not using SSL, and pidgin seems to be running without complaint. So deep42thought probably made the right diagnosis, that the problem is somewhere within nss.
<farkuhar>
Over in #crux-arm, sajcho suggested to build nss with clang as the compiler, to address an unrelated problem. I'll try that and see if it resolves FS#1933.
<beerman>
sounds like this should be an upstream ticket for libpurple/pidgin..
SiFuh_ has quit [Remote host closed the connection]
SiFuh has joined #crux-devel
farkuhar has left #crux-devel [#crux-devel]
farkuhar has joined #crux-devel
<farkuhar>
I just rebuilt nss using sajcho's Pkgfile (the one that compiles with clang). With pidgin linked against this build of nss, the "illegal instruction" errors of FS#1933 are not occurring. I agree with beerman that a ticket should be filed upstream, but maybe with nss or gcc, not pidgin/libpurple.
<beerman>
dunno, have you ran the test cases of nss with gcc yet?
<farkuhar>
not yet. I'm not really interested in filing the upstream ticket, as I'm not a regular pidgin user. deep42thought can write that ticket himself, or just use the nss-compiled-with-clang workaround.
<farkuhar>
anyway, before FS#1933 gets closed as a "wontfix", I suppose I could write a comment about the clang solution, in case deep42thought isn't reading the IRC logs.
<beerman>
yep, makes sense. thanks for taking care about it either way
<beerman>
clang may introduce other weirdness though, so just keep that in mind