I've been writing a lot of code that uses The GNU Multiple Precision Arithmetic Library, and I would like to be able to use gdb on these programs. The issue is being able to print the values of gmp variables. The documentation suggests using this .gdbinit, but if I try I tend to get segfaults.

So, I have two questions.

1. How would I use the .gdbinit file to print the value of a gmp variable?
2. I could try to call gmp_printf (which is in the gmp library) to print out the values, but these functions are unavailable to me (no symbol "gmp_printf" in the current context). How would I use these functions?

Thanks in advance.

Hi -

1. .gdbinit is provided as an add-on precisely to ALLOW you to print GMP variables from GDB. You're apparently already using it (that's why you're getting the segfaults). *Why* you're getting the segfaults - I couldn't even begin to speculate without looking at a stack trace and debugging the code (your application, GMP, and everything in between).

2. An alternative to using .ginit is to simply recompile GMP and make sure your GDB search path includes a) all your own code (compiled with "-g" or equivalent), and b) all the GMP code (also compiled with "-g" or equivalent). One problem with this approach, as mentioned in the documentation you cited, is that GMP has a number of symbols with the SAME NAME, but in PARALLEL directories. Which can be messy to deal with...

3. A third approach - one which NEEDN'T require gdb and/or .gdbinit - is to use the GMP call "gmp_printf()". You simply add this to your application code - just as you would add "printf()".

If you haven't already done so, I would strongly encourage you to join a GMP mail list and seek advice there, too:

http://gmplib.org/mailman/listinfo

'Hope that helps .. PSM

1 members found this post helpful.

Well, first I read Paul's answer - I am not sure what segfaults occur - in the program or in 'gdb' itself - I'm afraid the latter, but I am not sure.

A couple of thoughts not directly addressing your problem:

1) if your 'gdb' is not the newest (7.1) and it crashes, build the newest one from source - I've just done it using

commands - 'gdb' needs 'ncurses' (even though the documentation doesn't mention it), so you'll need development headers. Hopefully your stock gcc + ncurses will be good enough.

2) gmp-5.0.0 might be "odd", even though I didn't look inside. I mean that gmp-5.0.1 + mpfr-3.0.0 used to build gcc-4.4.4, which are all then used to build sbcl-1.0.39, cause build failure of the latter.

OTOH, gmp-4.3.2 + mpfr-2.4.2 used to build gcc-4.4.4, which are all then used to build sbcl-1.0.39, produce normal build.

I didn't dig deeper.

1 members found this post helpful.

I figured out what's wrong. It turns out that the .gdbinit file provided is not exactly correct; one cannot apply __gmpz_dump to $->_mp_num (or $->_mp_den); one must apply it to its ADDRESS for __gmpz_dump to get what it's expecting. D'oh.

I will join the mailing list for future concerns. Thanks to both of you for your comments.

1 members found this post helpful.

Cool - glad you got it working!

And thank you for sharing HOW you got it working!

Sincerely .. PSM