cygwin build_alias issue
bvacaliuc at ngit.com
Wed Sep 1 16:13:46 UTC 2004
I've been following this thread for the last week, and I wanted to share
some of my observations so far.
I also tried the strace, using '( strace --mask=minimal --output=strace.log
make VARIANT=DEBUG all ) | tee build.log'; the build processes executes to
completion every time, albeit slower. The log output is only 1.7M.
I would like to suggest that the issue is a localized resource recovery
problem. I.e. if during a pipeline a delay is introduced (in this case via
strace), the system recovers its released resources and subsequent commands
will not fail.
I have not found any references to controlling such resources. One thing I
did do is have the task manager open while the build was running. I saw
that the average number of processes increased by 18 while the build was
running. One windows service 'CSRSS.EXE' routinely took 20-40% of the CPU
for brief moments and increased its active handles (on average, by visual
accounting) by 40 during configure and 100 during build. The handles used
by CSRSS.EXE during build peaked at +137 from the nominal value (in my case
There is a M$ Knowledge Base article #126962 that may have some bearing on
I am going to apply the registry edits they specify and re-run the builds;
I'll let you know what my results are.
> -----Original Message-----
> From: Scott Newell [mailto:newell at cei.net]
> Sent: Wednesday, September 01, 2004 9:21 AM
> To: rtems-users at rtems.com
> Subject: Re: cygwin build_alias issue
> At 12:07 AM 9/1/2004 , Ralf Corsepius wrote:
> >But ... still nobody seems to have been able to "sh -x" or
> strace this
> >issue, ... so all this is wild guesses.
> Hmm, I thought I mentioned strace this weekend...anyway, I
> tried strace. Generated in excess of 10GB of log files, and I
> couldn't get it to give me the build_alias error a single time.
More information about the users