Re: [jamboost] Re: build.bat change to help debug bjam crashes

From: Beman Dawes (bdawes_at_[hidden])
Date: 2003-07-14 18:42:03

At 05:55 PM 7/14/2003, David Abrahams wrote:

>Beman Dawes <bdawes_at_[hidden]> writes:
>
>> At 02:53 PM 7/14/2003, David Abrahams wrote:
>> >Beman Dawes <bdawes_at_[hidden]> writes:
>> >
>> >> I'm still getting occasional bjam crashes which disappear if the 
same
>> >> command is given a second time.
>> >>
>> >> It just happened again, so I ran a "build vc7 --debug" to get a 
debug
>> >> build. But there was no crash with this debug executable.
>> >>
>> >> Unfortunately, the debug build executable is so slow that it isn't
>> >feasible
>> >> to use it for all runs until the bug is captured.
>> >>
>> >> Would it be possible to modify build.bat to create a custom 
bjam.exe
>> build
>> >
>> >> that does include debugging symbols, but otherwise is a release
>build?
>>
>> My guess is that would be fine.
>>
>> >Not quite easy, but I don't think you need optimization. Try 
checking
>> >out tools/build/jam_src again and building with --debug --noassert.
>>
>> Didn't seem to have the desired effect. I assume a -DNDEBUG was 
expected.
>
>> Below is the command line (-d2).
>
>Sorry, please try again.

OK, -DNDEBUG appeared on the command lines. Thanks.

I'm testing now. The initial phase seems very slow - apparently 
optimization is more important than I would have thought. But at least we 
stand a chance of finding the pesky bug that has been a bother for a long 
time.

It is probably worth noting that V1 is performing quite well, and has been 
for quite a while now. Overall, Boost.Build for regression testing is now a 
great success story IMO.

Thanks,

--Beman