| 1 |
15. Nov 2017 23:54 by agl@×××××××.au: |
| 2 |
|
| 3 |
|
| 4 |
> On 16/11/17 11:05, > tuxic@××××××.de> wrote: |
| 5 |
>> Hi, |
| 6 |
>> |
| 7 |
>> building firefox 57.0 failed on my system - it looks like |
| 8 |
>> the last stage (linking) fails. |
| 9 |
>> |
| 10 |
>> I attached the build.log to this mail. |
| 11 |
>> |
| 12 |
>> Is there a way around this? |
| 13 |
>> |
| 14 |
>> Cheers |
| 15 |
>> Meino |
| 16 |
>> |
| 17 |
> |
| 18 |
> |
| 19 |
> First thing I do when I have a problem with one of the larger apps, |
| 20 |
> Firefox, Thunderbird, LibreOffice is enable the make feature, in |
| 21 |
> make.conf, that keeps temp files around: |
| 22 |
> |
| 23 |
> FEATURES="ccache keeptemp keepwork candy" |
| 24 |
> |
| 25 |
> and then set the make options so that only one thread is doing stuff: |
| 26 |
> |
| 27 |
> MAKEOPTS="-j1" |
| 28 |
> |
| 29 |
> I just sometimes find that the build system gets a bit confused, with |
| 30 |
> multiple threads, and the one thread "straightens" things out. |
| 31 |
> |
| 32 |
> But then again, this might be a bug and I have no idea as to what I'm |
| 33 |
> talking about..... |
| 34 |
> |
| 35 |
> Andrew |
| 36 |
|
| 37 |
|
| 38 |
|
| 39 |
|
| 40 |
if running just one thread fixes it, it's probably a race condition, i.e. the multiple threads aren't on the same page all the time. there's multiple processes and they aren't wiating/notifying each other and one of them therefore gets data the other process is still chainging and there's no "interlock" to let the other process know or make them wait for eachother. obviously it's a bug of some sort and is hopefully being worked on. having said that, compilers are very, very complicated, especially with "optimization" and RISC processors. might also be a sort of cache incoherentcy. |
| 41 |
|
| 42 |
|
| 43 |
|
| 44 |
|
| 45 |
|
| 46 |
mad.scientist.at.large (a good madscientist) |
| 47 |
-- |
Archives