Re: [Boost-build] Early bjam abort

Subject: Re: [Boost-build] Early bjam abort
From: Steven Watanabe (watanabesj_at_[hidden])
Date: 2012-04-23 18:32:16

AMDG

On 04/23/2012 02:37 PM, Nogradi, Chris wrote:
> I have a project with ~70k targets including generated header and source files that builds fine without concurrency (-j1):
> 
> ...found 72133 targets...
> ...updating 5981 targets...
> <snip>
> ...updated 6237 targets...
> 
> but stops in the middle of the build when using concurrency (-j2 to -j8). There are no failures reported, it just stops and reports the target count that was updated but does not build all targets:
> 
> ...found 72133 targets...
> ...updating 5981 targets...
> <snip>
> ...updated 5930 targets...
> 
> If I run it again a number of times w/ concurrency enabled it will eventually complete (not sure why target count increased):
> 

The target count probably increased because
it includes targets that were found by scanning
generated headers.

> ...found 78019 targets...
> ...updating 2094 targets...
> <snip>
> ...updated 1778 targets...
> 
> ...found 78019 targets...
> ...updating 2088 targets...
> <snip>
> ...updated 1772 targets...
> 
> ...found 78019 targets...
> ...updating 316 targets...
> <snip>
> ...updated 316 targets...
> 
> Any ideas on how to debug this issue?
> 

-d+13 might help.
Do the targets that get skipped silently have anything in common?

I'm guessing that the problem is that rescanning
generated headers is messing up asynccnt for
some targets.

> I am using SVN 77778 but have seen the issue since 4-11-2011 (not sure what SVN version at that time).
> 

In Christ,
Steven Watanabe