On 2010-06-07 at 08:53 -0400, John Jetmore wrote:
> The machine I've been using to work on the test suite is a RHEL 4
> machine. I tried to build from HEAD this morning to see what havoc
> Phil wrought on the rules but I can't get past buildconfig. It
> compiles but segfaults when running. I tried it on tahini to rule out
> a 32b v 64b issue as well as something purely local and it segfaults
> there also.
>
> Phil, can you take a look at this? I will check back tonight and work
> if it's still an open issue but I'm not going to dive in this morning.
Works on FreeBSD 7 / amd64 and on a Linux i686 laptop where the OS is
based on Ubuntu 8.04.2.
There were two CLs affecting buildconfig:
* don't build as root
* conversion warnings cleanup
(assumption that single-long was 4 bytes)
Diff:
http://vcs.exim.org/viewvc/exim/exim-src/src/buildconfig.c?r1=1.15&r2=1.17
I do see that I left disabled the PRIdMAX usage which should be the
default on modern OSes (when not C99) -- sorry, I was testing the other
logic worked without warnings. I'll fix that.
At an eyeball, none of the changes look like they could cause a crash.
My big concern was including <inttypes.h> on older platforms.
To narrow it down: does including -std=c99 in CFLAGS deal with the
problem? (Not proposing as a solution, but if this does deal with it
then the problem is in the SIZE_T_FMT code).
-Phil
