On 06/27/2011 07:02 PM, David G. Koontz wrote: >> Your test case is defective. > I've found no evidence of a bug in ghdl. It seemed worth checking why a > contraindication was found during investigation (num to numeric_std). I > figured reporting it might prevent any one pointing out the apparent > inconsistency in my posts, the original comment left the issue dangling.
You are right -- the test case I posted was made too hastily, is broken, and doesn't demonstrate any GHDL misbehavior. Sorry for the wild goose chase. However, I'm *quite* confident that problem I was seeing was definitely a GHDL bug. Not only does the code work in ActiveHDL and Xilinx ISIM but fail in GHDL, but it starts working in GHDL when I switch from buffer to out, and adding a bunch of asserts sort of caught it in the act. Unfortunately I can't post the original code as it's not mine, and it obviously doesn't help much to make claims without a reproduction recipe. I've already worked around this for the time being (and it was a pain to try to narrow down in the first place) but if I have time or run into this again, I'll try to make a test case that actually reproduces the problem I saw. _______________________________________________ Ghdl-discuss mailing list [email protected] https://mail.gna.org/listinfo/ghdl-discuss
