| ginggs | xnox: re LP: #1950283 you disabled LTO completely there, but it seems upstreams supports LTO, and it only disabling it for that one test seemed to work https://launchpad.net/~ginggs/+archive/ubuntu/testing/+sourcepub/12880546/+listing-archive-extra | 12:35 |
|---|---|---|
| ubottu | Launchpad bug 1950283 in r-cran-rsvd (Ubuntu) "r-base: gfortran and gcc disagree on double complex" [Undecided, New] https://launchpad.net/bugs/1950283 | 12:35 |
| ginggs | and the autopkgtests of the affected packages succeeded | 12:36 |
| ginggs | i ask because there's a new upstream version of r-base in debian i think we should sync/merge before enabling autosync | 12:37 |
| xnox | ginggs: sounds good. | 14:20 |
| xnox | ginggs: yeah, it's always nice to do things like that before autosync | 14:20 |
| ginggs | xnox: well yes :) but the question (which i didn't actually ask) was does disabling LTO for that config test seem sane? and how to propose that to upstream? | 14:24 |
| ginggs | it is a bit weird, it only seems to happen with -O2 and LTO | 14:25 |
| ginggs | -O1 or -O3 with LTO is fine | 14:27 |
| xnox | ginggs: that is interesting, it would be nice to isolate the conftest.c itself; outside of all the configury things. And see if we can somehow improve it to work correctly. Cause now it sounds that said conftest is lto buggy and doesn't link/reference things it tries to use. | 14:54 |
| ginggs | xnox: i'll mail it to you | 14:55 |
| xnox | ginggs: or subscribe me to the bug report where you attach it? =) | 14:57 |
| ginggs | xnox: attached to LP: #1950283 (reduce, reuse, recycle) | 16:39 |
| ubottu | Launchpad bug 1950283 in r-cran-rsvd (Ubuntu) "r-base: gfortran and gcc disagree on double complex" [Undecided, New] https://launchpad.net/bugs/1950283 | 16:39 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!