Re: Re: cvs: php-src / Makefile.gcov NEWS configure.ingen_php_cov

Hi Marcus. Thanks for reviewing the patch! I agree with you that using environment vars is excessive in this script. The only reason I've used it is to allow the user to pass arguments to valgrind (like the suppresions file). After reading the valgrind manual, I've discovered that it is also possible to pass the options, using the VALGRIND_OPTS environment var. I've updated the patches: http://mega.ist.utl.pt/~ncpl/php_tests_valgrind.txt http://mega.ist.utl.pt/~ncpl/php_tests_valgrind_w.txt Maybe a 'make test_leaks' could be available as well? :) Regards, Nuno ----- Original Message -----

Hello Nuno, the use of environment is inacceptable here. Just invent some new switch. Probably "run.tests.php -m" as in memleaks? Also defining VALGRIND and then doing 'system_with_timeout(VALGRIND.$cmd);' is ugly style and there is no need to have that define anyway. You could easily just change $cmd. marcus Saturday, October 22, 2005, 2:09:08 AM, you wrote:

Hi Marcus,

And what about generate the files with *relative* paths. The #line directives are very usefull for debuging, so I think they should continue there, but with relative paths, not to the path where Ilia or Jani have their php sources.

I've updated my patch (to latest HEAD) to 'run-tests.php' to allow the use of valgrind. diff -u: http://mega.ist.utl.pt/~ncpl/php_tests_valgrind.txt diff -u -w: http://mega.ist.utl.pt/~ncpl/php_tests_valgrind_w.txt (much clear to see what I've change)

I would like to ask somebody to review it.

Thanks, Nuno