On 19.02.2020 23:50, Joerg Sonnenberger wrote: > On Wed, Feb 19, 2020 at 11:29:09PM +0100, Kamil Rytarowski wrote: >> On 19.02.2020 23:03, Joerg Sonnenberger wrote: >>> On Wed, Feb 19, 2020 at 04:10:17PM +0100, Kamil Rytarowski wrote: >>>> On 19.02.2020 14:32, Joerg Sonnenberger wrote: >>>>> Module Name: src >>>>> Committed By: joerg >>>>> Date: Wed Feb 19 13:32:40 UTC 2020 >>>>> >>>>> Modified Files: >>>>> src/external/apache2/llvm/config/llvm/Config: config.h.in >>>>> >>>>> Log Message: >>>>> Revert last. It was objected to pre-commit, there is no actual error >>>>> analysis or report and there is existing logic supposed to handle this. >>>>> >>>>> >>>>> To generate a diff of this commit: >>>>> cvs rdiff -u -r1.2 -r1.3 \ >>>>> src/external/apache2/llvm/config/llvm/Config/config.h.in >>>>> >>>>> Please note that diffs are not public domain; they are subject to the >>>>> copyright notices on the relevant files. >>>>> >>>> >>>> Objected pre-commit? Nothing reached me and I see nothing in logs. >>>> >>>> No report? It was discussed in public syzkaller ML and semipublic IRC >>>> channel with 5 people involved. >>>> >>>> https://syzkaller.appspot.com/text?tag=CrashLog&x=11aafc09e00000 >>>> >>>> No analysis? There is analysis in the commit message that we remove >>>> optional features. >>>> >>>> I request either to reintroduce it now or land another fix. >>> >>> While I am on many mailing lists, I'm certainly not on all lists. I >>> answered in the only forum where I saw the message. I am strongly >>> objecting here because you make the tools compiler a lot less usable. >>> You haven't acutally shown *why* the existing logic doesn't work. >>> >> >> The existing logic does not work as it is breaking bootstrap on >> Linux/Ubuntu as presented here: >> >> https://syzkaller.appspot.com/text?tag=CrashLog&x=11aafc09e00000 > > Yes and that is completely useless for figuring out the why. > > Joerg > We don't have any direct reproducer (we tried) and we must figure it out from syzkaller bot. We don't have access to the machine and a limited access to an admin over there (who has no expertise on BSDs). This failure breaks kMSan build for long time now. There is full build log. Calling syzbot report useless does not help. Lack of support in the build phase will result in removal of the bot instance and turn our work into futile effort.
Attachment:
signature.asc
Description: OpenPGP digital signature