Comment # 10 on bug 1202027 from Richard Biener

(In reply to Martin Li������ka from comment #9)
> > At least I can't find
> > a good reason why it should not?
> 
> For instance, a temporary file name can have a different length and cause a
> bit different memory fragmentation, but who knows..

I think mktemp & friends produce consistent lengths (good) but just different
characters (sometimes problematic).  Of course filesystem activity might
need different amounts of memory internally but that's hopefully released...

> Lemme try the non-LTO buids..

Those should at least be easier to analyze - I'm crossing my fingers ;)