windows 64bit build failures, not always, randomly failing

Magnus Ihse Bursie magnus.ihse.bursie at oracle.com
Fri Sep 7 01:28:58 PDT 2012


On 2012-09-06 23:48, Kelly O'Hair wrote:
> Just an FYI...
>
> I seem to be getting the Windows 2003 X64 failures running /usr/bin/grep:
>
> ...
> /usr/bin/grep -e '\.class$' /cygdrive/c/jprt/T/P1/203843.jprtadm/s/build/windows-x86_64-normal-server-release/jdk/lib/_the.jars.contents >
>   /cygdrive/c/jprt/T/P1/203843.jprtadm/s/build/windows-x86_64-normal-server-release/jdk/lib/_the.rt.jar.contents.tmp
> /usr/bin/mkdir -p /cygdrive/c/jprt/T/P1/203843.jprtadm/s/build/windows-x86_64-normal-server-release/jdk/lib
> /usr/bin/rm -f /cygdrive/c/jprt/T/P1/203843.jprtadm/s/build/windows-x86_64-normal-server-release/jdk/lib/_the.resources.jar.contents
>   /cygdrive/c/jprt/T/P1/203843.jprtadm/s/build/windows-x86_64-normal-server-release/jdk/lib/_the.resources.jar.contents.tmp
> make[3]: *** [/cygdrive/c/jprt/T/P1/203843.jprtadm/s/build/windows-x86_64-normal-server-release/jdk/lib/_the.rt.jar.contents] Error 1
> make[3]: *** Waiting for unfinished jobs....
>
> The grep:
>
> /usr/bin/grep -e '\.class$' _the.jars.contents > _the.rt.jar.contents.tmp
>
> is failing with an Error 1.
>
> Found this reference:
>       http://cygwin.com/ml/cygwin/2009-05/msg00036.html
>
> Yipes... grep can fail if the grep console output is more than 64K???  :^(

Not really; that was fixed in 2009. Read on in the thread: 
http://cygwin.com/ml/cygwin/2009-05/msg00120.html

Nevertheless, I've seen strange problems myself on Windows. Twice, when 
I started digging into some hangs, I found the hanging command to be a 
find:
+ /usr/bin/find 
/cygdrive/c/localdata/hg/build-infra-jdk8/build/windows-x86-normal-server-release/images/c 
-type f -a '!' -name '_the.*' '(' -name FALSE_DUMMY -o -name '*.java' -o 
-name '*.c' -o -name '*.h' ')'

Process explorer stated that this process was "suspended", which is a 
bit weird. In Windows that's a debug state which a process should not 
normally reach. Resuming it didn't have any effect though (not 
surprising). My guess is that it's some kind of half-disabled assert in 
the cygwin code.

/Magnus



More information about the build-infra-dev mailing list