Re: BUG in 3.80 (virtual memory exhausted, xrealloc, variable_buffer_output)

[Dr. Jörn von Holten]

patch_item_1022.patch seems to solve the bug! Hopefully.

Dr. Jörn von Holten wrote:

> Hi,
>
>   I reduced my problem to the following short Makefile
>
> -----------------snip---
>
> define A
> 12345: $(1:.idl=.hh) $(1:.idl=S.h) $(1:.idl=C.h) $(1:.idl=SK.cc) 
> $(1:.idl=DynSK.cc)
> endef
>
> $(eval $(call A,01234567890123456789012345678901.idl))
>
> ---snap-----------------
>
> which, when evaluated (e.g. "make -r") with 3.80 results in a
>
> *** virtual memory exhausted.
>
> I traced the problem into the sources, which I saw the very 1st
> time, and what I found is:
>
> A call to variable_buffer_output results in a xrealloc call when
> appending the final '\0' while doing the eval. It seems as if the
> new memory position of the variable_buffer is not correctly passed
> back then as the next call to variable_buffer_output then works on
> an illegal pointer.
> This is just a guess and I have no time to go deeper into those sources.
> Maybe it is even a wild-overwrite.
>
> This problem occurs at least on Linux and on Solaris so
> it does not seem to depend on a certains OS.
>
> cheers
> Jörn

--
Dr. Jörn von Holten
Senior Consultant

Barco Orthogon AG
Hastedter Osterdeich 222
D-28207 Bremen
Tel +49-421-20122-426
Fax +49-421-20122-999
www.barco-orthogon.com
address@hidden