On Thu, 2014-01-09 at 22:27 -0500, Adam Jensen wrote: > On Fri, 10 Jan 2014 03:58:06 +0100 (CET) > [email protected] wrote: > > For FreeBSD, I'd like to see this issue fixed, but I don't really know > > how to help. > > > > I will start tinkering with the test-suite this evening. I've been > thinking about writing a testing driver script that will launch > [parallel][1] jobs and go through the entire test-suite, regardless of > any particular failures, then assemble a nicely formatted report.
I also have an effort under way on this (which was superseded by some of Tristan's commits last week) but think I have observed the same failure (gcc build, Linux : Debian Jessie). yes: Test: 375 ghdl -a --std=93c vhdl-93/ashenden/compliant/ch_13_fg_13_26.vhd vhdl-93/ashenden/compliant/ch_13_fg_13_26.vhd:33:20: port "a" already associated in primary binding and three similar offences. I have a modified test script allowing the full set of (analyze / elab / run / analyze_failure / elab_failure / run_failure) operations, and some modified .exp scripts running tests that currently only analyze, but didn't want to pollute the source tree with a half-baked known-faulty test suite before 0.31 went out. I'd like to pursue this further. I ran into the test sequencing problem Tristan described, and a couple of others and would like to make more progress before committing (and breaking the testsuite in /default). Happy to share them privately, or commit them then roll them back so they are accessible, if you want. As Tristan says it's not too slow to run single-threaded (though with gcc, more like 5-10 mins than 2 mins), and there are sequence dependencies. So if you wish to parallelise, I recommend doing so at the directory level rather than at the file level : there will be no dependencies or ordering there. - Brian _______________________________________________ Ghdl-discuss mailing list [email protected] https://mail.gna.org/listinfo/ghdl-discuss
