On 11/29/18 12:35 AM, Brooks Moses wrote: > Fair enough -- but having done cross-testing on both GCC (with > DejaGnu) and glibc (without it), and having also done some substantial > porting work to get the GCC testing working with our systems, I have
I'll note that GCC testing is much easier then GDB testing. :-) Much of the difficulty in DejaGnu was to support interactive applications. That and Canadian cross testing of remote windows based cross compilers. > One of the interesting things I learned from that is that I actually > ended up doing almost no porting of the DejaGnu code itself; the > porting was almost all of functionality provided by the "library" code > that the GCC testsuites add on top of it. Admittedly, that's because > the "figure out which tests to run" and "run a test on this kind of > machine" functionality is provided by our build-farm infrastructure, > but I still found it noteworthy how separable the pieces were. Well, that was actually a design decision. :-) DejaGnu is a framework for creating a testsuite, the runtest program is just a utility. The last build farm I built just ran the entire testsuite for each patch to make sure we didn't break anything, so pretty simple. - rob - _______________________________________________ DejaGnu mailing list DejaGnu@gnu.org https://lists.gnu.org/mailman/listinfo/dejagnu