[ale] gcc memory leak?

Danny Cox danny at compgen.com
Mon Oct 16 13:41:02 EDT 2000


On Mon, 16 Oct 2000, Wandered Inn wrote:

> Had a scary problem with a compile yesterday, and based on what
> happened, I'd expect that there might be some kind of memory leak in
> gcc.  I'm looking for some insights.  I'm running 2.95.2.  I was
> compiling the latest version of sane, the scanner package.  When it got
> to the point that it was compiling the final backend, umax scanner
> stuff, I could see my memory usage quickly go up.  Once all the memory
> was gone, then the swap usage began to climb.  The first time this
> happened, I wasn't aware what caused it as I was running a number of
> packages. I was too late to do anything about it and it locked my box
> up.  After a hard reboot, I fired off the compile again and
> intentionally kept an eye on the memory usage.  The problem is very
> consistent.  I can kill the compile and the memory usage drops
> dramatically.
> 
> My point is, a compiler shouldn't do this.  I would not expect when
> attempting to compile a piece of code, that gcc could cause this kind of
> problem.

	I've read that the optimization pass does this.  If you will
attempt to compile this one module with no optimization, it should succeed.
It may well be a bug, but when it comes to optimizers, it's all black magic
to me.

Danny

--
To unsubscribe: mail majordomo at ale.org with "unsubscribe ale" in message body.





More information about the Ale mailing list