[ARM] Regression between r215896 and r215960

[Jonathan Roelofs]

On 8/20/14, 1:39 PM, Renato Golin wrote:
> On 20 August 2014 14:23, Renato Golin <renato.golin at linaro.org> wrote:
>> 846 and 845 are both bad, and ever since it first regressed, 20 new
>> failures have been creeping in... :(
>
> So, apparently, by going backwards (via bisection) and not cleaning, I
> made all good commits fail, so there is some level of caching
> involved. Though, I cleaned the bot (even the source), and it still
> fails, so it's not *just* caching.
I have noticed that for auto-fu incremental re-builds, the compiler_rt build is 
missing the dependency on the just-built clang, so it doesn't get re-built if a 
new clang is made. Maybe this is contributing to what you're seeing?
>
> I'm testing clean sequentially now, and each hop takes about 1h20min
> to finish, which is really fast for a full bootstrap on ARM hardware.
Still sounds painful :(

I wonder how much disk space would be eaten up by saving every build always?


Cheers,

Jon
> I'm keeping all the logs and will let you guys know when it spots
> something.
>
> cheers,
> --renato
> _______________________________________________
> cfe-commits mailing list
> cfe-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
>

-- 
Jon Roelofs
jonathan at codesourcery.com
CodeSourcery / Mentor Embedded