[Gc] GC_unix_get_mem on IRIX5

Boehm, Hans hans.boehm at hp.com
Fri Apr 21 17:01:24 PDT 2006


Thanks.

You are correct about the bug.  Good catch.  I put a patch in my source
trees, from where it will eventually propagate out.

I don't think that USE_MMAP is necessarily used on Irix.  But to reach
either of those paths, you

a) need to be running a recent Irix version,
b) run out of memory, and
c) have at least one other unlikely event occur.

If someone indeed has ever seen a failure, they probably attributed it
to the fact that recovery from out of memory conditions usually isn't a
sure thing anyway, since things other than sbrk may fail.  I'm not
surprised this was only found by code inspection.

Hans

> -----Original Message-----
> From: gc-bounces at napali.hpl.hp.com 
> [mailto:gc-bounces at napali.hpl.hp.com] On Behalf Of 
> Jean-Baptiste Nivoit
> Sent: Friday, April 21, 2006 4:34 PM
> To: gc at napali.hpl.hp.com
> Subject: [Gc] GC_unix_get_mem on IRIX5
> 
> I just stumbled upon the following when reading the code in 
> os_dep.c around line 1630 ( not RS6000, not USE_MMAP ):
> when the IRIX5 #define is on, i see 2 exits points from that 
> function that fail to release the sbrk mutex, which would 
> prevent other threads from using sbrk..
> 
> I guess on IRIX5 the USE_MMAP option is used, which is why 
> nobody saw this as a problem until now (I don't have such a 
> system, i'm just inferring this from reading the source code).
> 
> Feel free to correct me if i'm mistaken.
> 
> 	jb.
> _______________________________________________
> Gc mailing list
> Gc at linux.hpl.hp.com
> http://www.hpl.hp.com/hosted/linux/mail-archives/gc/
> 



More information about the Gc mailing list