| kees | doko: so, in comparing testsuite output between Debian and Ubuntu, I notice that there some failures on Debian too, so I was going to skip those. | 07:14 |
|---|---|---|
| doko | kees: yes, that's ok. for upstream inclusion, prepare these tests for trunk/4.4 ... | 07:17 |
| kees | doko: okay, sounds good. I'm worried they will ignore a lot of them that add things like "-Wno-format" to dg-options. Any thoughts if they'll accept those kinds of things? | 07:18 |
| kees | also, I'm curious about one part I haven't been able to sort out yet. Part of the testsuite does not honor dg-options, so it requires a .x file, but something seems to be removing it before the test. (I added a .x and it shows up after debian/rules patch, but not after debian/rules build) Seen anything like that before? | 07:20 |
| kees | and, as far as stuff that doesn't pass in Debian, this one tries to allocate an insane about of memory (and fails): | 07:22 |
| kees | gcc.c-torture/compile/limits-blockid.c | 07:22 |
| kees | you've swapped out. :) | 07:23 |
| kees | oh, nm about the ".x" file, I just figured it out. :) still curious about upstream attitudes. | 07:24 |
| kees | doko: the gcc build runs the testsuite normally and with -fno-stack-protector. the latter is breaking the ssp tests since the "no" option is overriding the dg-options. Since there are no differences (besides ssp) between the two testsuite runs, can we just drop the -fno-stack-protect run? (it'd speed up builds too) | 16:01 |
| doko | kees: away now, will answer tomorrow or Monday | 16:01 |
| kees | doko: okay | 16:01 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!