On Sun, Mar 07, 1999 at 02:44:49AM -0800, Brian Warner wrote: > [EMAIL PROTECTED] (Sven LUTHER) writes: > > > so far i have compiled glib/gtk and imlib, but more will come, i just got > > all sources packages today. ... the glib/gtk1.2 are ok, in fact they will > > go to unstable once the source comes out of incoming ( i guess it will be > > built by the built daemon, since it worked fine for me.) > > Is that gtk+ built with the egcs-bug workaround? (the test is to run the > 'gtk/testgtk' program that is built along with the libraries, and choose the > 'clist' option: it will crash without the workaround). There's a bug in egcs > in which functions with a large number of arguments (maybe 8 or 9?) compiled > with -fPIC -O2 will crash. Compiling such files -O1 is OK. The only files I've > ever seen trigger this bug are in gtk+. I reported the bug to the egcs folks > last year and it's in the testsuite (980523-1.c), but as of 2 weeks ago it had > not yet been fixed. > > The workaround for gtk+ is to build with something like this: > > CFLAGS=-O2 ./configure > cd gtk; make CFLAGS=-O1 gtkclist.lo gtkctree.lo; cd .. > make > > This may be a problem for a while.. if egcs hasn't been fixed by now, I > suspect the bug must be pretty tough. > > You can also build that testgtk after the fact to test for the problem: > tar xzf gtk+-1.2.0.tar.gz > cd gtk+-1.2.0/gtk > gcc -o testgtk `gtk-config --cflags` testgtk.c `gtk-config --libs` > ./testgtk > [hit "clist"]
Ok nice you tell me this, i will test this stuff nextly ... IT is a ppc only bug i guess ? also i have been experiencing gtk1.2/Xserver crashes with the color_selection widget, i thought this was related to me running X on top of planar amifb, but could it also be a egcs bug ? i will post a bug report with strace and other stuff a bit later ... Friendly, Sven LUTHER
