Re: [1.7] BUG: heap_chunk_in_mb=1536 breaks expect.

[Corinna Vinschen <corinna-cygwin at cygwin dot com>]

On Mar 15 22:53, Dave Korn wrote:
> 
>   Apologies for what is going to be quite light on details as bug reports go,
> but I've only just noticed it and I haven't got a simple testcase yet.
> 
>   As the subject line says, heap_chunk_in_mb=1536 breaks expect.  The
> situation that I'm seeing this in is when attempting to run the binutils
> testsuite.
> 
>   With heap_chunk_in_mb set to 1536, "make -k check" gets as far as the very
> first *.exp file in the binutils testsuite (binutils-all/ar.exp), and then
> sits there, spinning its wheels and eating CPU.  It doesn't happen if
> heap_chunk_in_mb is set to 1024, 1280, 1408, 1520 and the test run completes.
>  It does happen at 1535, and 1532.  At 1528 to 1530 it gets a few testcases
> further into the run before stalling.  Probably the exact cut-off point is
> environment-dependent.
> 
>   Sorry not to have much information, but at least there's a simple
> workaround: turn it down a bit.  If I find out more I'll report back.

The code handling heap_chunk_in_mb hasn't changed at all since 1.5.25,
except for the registry location where it's loaded from.  I'm wondering
if the mere size of the heap chunk is enough here to disturb
functionality which requires more memory in 1.7 than in 1.5.  For
instance, practically all file operations need more memory now (buffers
for conversion to UNICODE, enough room for long paths).  And 1.5 Gigs +
sizeof all DLLs + sizeof executable is quite easily filling up the 2
Gigs virtual mem size for the process.


Corinna

-- 
Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Project Co-Leader          cygwin AT cygwin DOT com
Red Hat

--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/