Re: [jamboost] bjam crash information

From: Beman Dawes (bdawes_at_[hidden])
Date: 2003-07-07 11:21:52

At 11:05 AM 7/7/2003, Vladimir Prus wrote:

>Beman,
>
>> I added an assert:
>>
>> if (t->flags & T_FLAG_INTERNAL) {
>> assert( t->original_target );
>> t = t->original_target;
>> }
>>
>> /* Clean current includes */
>> if (t->includes) {
>> t->includes = 0;
>> }
>>
>> The assert does fail, after around 30 seconds of 100% CPU activity.
>
>Good news first. I've just comitted a fix for this crash. It turned out,
>T_FLAG_INTERNAL was #defined to the same value as T_FLAG_RMOLD, and since 

>test system uses RMOLD, everything broke.

Thanks for the fix.

>The bad news is that it's very new code, so the crash you've had before
>must be caused by something else.

Well, at least now I'm all setup with a debug build ready to test with. If 
anything else surfaces I should be able to give a better error report than 
just "it broke."

>BTW, running bjam in status directory takes a lot of time. Is it normal, 
or
>it was faster previously?

My debug build takes about 40 seconds for a particular test case. The same 
under the release build from jam_src/build.bat takes about 6 seconds.

A badly fragmented disk can cause these speeds to degrade very badly. And 
running the regression tests regularly tends to cause a lot of 
fragmentation.

--Beman