Re: [pcre-dev] Here is pcre-7.1-RC2 for you to play with

On Mon, 26 Mar 2007, Sheri wrote:

> > I wonder if the --disable-stack-for-recursion might be influencing the
> > malloc things? Have you tried without that?
> >
> >
> >
> Hi Philip,
>
> No, but have you tried compiling with it?

Yes. Previously I always did a check of that by hand, and now my
standard set of tests, which I have incorporated into a script, includes
tests with --disable-stack-for-recursion. However, I see that that
particular test also includes --disable-shared. I'll add a test with
exactly your configuration, and see what happens...

It compiles and runs the basic tests without any problems. This is on
Linux, of course.

> We disable stack recursion because otherwise, when pcre needs more
> stack space than allocated in the library, the program crashes. Prior
> to adopting the no-stack approach, some of the tests in pcretest
> crashed.

Ah, OK, you are living in a small-stack world. The tests run OK on Linux
with an 8Mb stack.

> Never saw these size allocation errors in pcretest results with 7.0.

That confirms my guess that it is something to do with the way things
are built and linked. The basic C code has not changed from 7.0, apart
from a few very small bugfixes (a line or two of code each - nothing of
any size).

If I really have to, I'll write you a small test program (or more
likely, a patch to pcretest to make it write out a lot more information
about what it is doing) so that we can try to track this one down.

> Apparently until the Cmake process is ready, my ability to test is going
> to be sorely limited. I tried replacing the c source files in a fresh
> pcre-7.0 dir with the ones from pcre-7.1RC2 to see if I could build
> pcre.dll and pcretest.exe the way I used to, but it didn't work.

That is probably because of the reorganization of the way some of the
source files are used - and some of them were renamed. Though I would
have expected it to work if you just replaced every .c and .h file in a
7.0 distribution with an identically named one from 7.1-RCx.

> BTW, you might want to tell that person writing to the web site to look
> at the sample program and the google programs.

Good point.

Philip

--
Philip Hazel, University of Cambridge Computing Service.