debugging a cygheap leak

[Christopher Faylor]

On Thu, May 01, 2008 at 06:23:35AM -0600, Eric Blake wrote:
> The m4 development head has a test where it repeatedly loads and unloads a 
> dynamic module using libtool to wrap dlopen() and friends.  This is working 
> fine under 1.5.x, with no memory growth after the first reload. But in 
> 1.7.0, memory usage continually grows, then the test is crashing after 
> about 1700 iterations with a call to api_fatal,
>
>      1 [main] m4 18392
> \\?\C:\cygwin\home\eblake\m4-head\build\src\.libs\m4.exe: *** fatal
> error - cmalloc would have returned NULL
>
> Any ideas on how to go about debugging this?  It looks like a memory leak 
> regression in 1.7.0.  But with my self-built cygwin1.dll and cygwin1.dbg, I 
> don't know where to set a breakpoint to catch it - trying 'b api_fatal' 
> fails.  Meanwhile, I'll continue trying to come up with a STC that doesn't 
> involve the entire m4/libtool wrapper around dlopen.

Build the DLL with --enable-debugging and CFLAGS=-g, if you can isolate
the test to just one invocation of m4 then:

set CYGWIN_DEBUG=m4

and do a backtrace from there.

cgf