Fail to produce dumps with big memory footprint processes (Linux)

What steps will reproduce the problem?
1. Run a large process with 130Gig VM size (in a server with 256 Gig RAM).  Use 
the ExceptionHandler function as described in the wiki:
http://code.google.com/p/google-breakpad/wiki/LinuxStarterGuide#Integrating_Brea
kpad_into_your_Application
2. Send sginal to initiatied dump (kill -11 <pid>);
    dumpCallBack() is called, but no file produced.
3. Reduce the VM size of the process to less than 100 Gig; Rerun process and 
dump file produced.

What is the expected output? What do you see instead?
dump file produced;  dumpCallBack() print the file path/name, but no dump file 
found.

What version of the product are you using? On what operating system?
R972; Linux

Please provide any additional information below.

It seems that even with in-process dump, it uses sys_clone which clone the same 
size of the failed process.  Perhaps some limitation is hit.  Related Issue 
420:  Can't clone because of no enough memory under linux 
http://code.google.com/p/google-breakpad/issues/detail?id=420 .  

In that case resources seem to be critically short; but in my case I should 
have enough resources to generate a dump file.

Again, aside from Chrome integration requirement, what it take to NOT fork a 
process and/or use CLONE_VM to reduce the use of memory?

Thanks.

Original issue reported on code.google.com by brjerom...@gmail.com on 21 Jun 2012 at 6:07

alrightlook / google-breakpad

Fail to produce dumps with big memory footprint processes (Linux) #486