Re: GNU Make Errors in Windows Application Event Log

[B G Ramprasad Torati]

Hi,

I've included the patch for job.c provided and built a debug version
of gnumake.exe. Now I am issueing a new test build using this debug
version of make. I want to catch as much information as possible with
this test build. One suggestion from JG is set
'first-chance-exceptions' to all break into the debugger.
How to set 'fist-chance exceptions'? Can somebody please add full
instructions for setting (and for restoring!) the "first-chance
exceptions".

Any other suggestions?

Thanks,
Ramp


On Tue, 02 Nov 2004 10:18:33 +0100, Alessandro Vesely <address@hidden> wrote:
> B G Ramprasad Torati wrote:
> > [...]
> > The build takes about 7 hours on a fast
> > machine and involves in building about 7000+ components (i.e. atleast
> > those many make.exe invocations), but reporting only around 250
> > exceptions with make.exe. That means not all the make.exe calls
> > reporting the problem.
> 
> I recall that running processes via temporary files is weak in 3.80.
> In particular, it does not guarantee temporary file name uniqueness.
> See patch
> <https://savannah.gnu.org/bugs/download.php?item_id=9748&item_file_id=1523>
> 
> I cannot guess how overwriting an existing temporary file may result
> in a read exception at 3f. Also, I'm not sure how cmd.exe would react
> to tampering a BAT file which is possibly being executed. However, I'd
> reckon that under such loads on a multiprocessor machine the chances
> to incur in that bug are high. (Details depend on the implementation
> of tempnam().)
> 
> If possible, I would try with a built using job.c from current CVS.
> 


-- 
Cheers..
Ramp