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

Paths can be made relative (at least re2c parsers, I don't know about flex/bison). This would just need a tweak in the makefile, to cd to the folder where the 'file.re' is, instead of calling re2c with an absolute path. I'm sure Jani can do this easily. Without this, lcov stops in the first parser, thus breaking the coverage report. Nuno P.S.: btw, the cvs version of lcov has a nice --show-legend option :) ----- Original Message -----

If paths could be made relative, that's the ideal solution. Otherwise I agree with Marcus that when --enable-gcov is on the ./configure line we should regenerate the parsers and scanners to omit the debugging information. Perhaps it would be wise in this circumstance to also tack on a little warning stating this has occurred just to keep people from complaining. John On Fri, 2005-10-21 at 20:08 +0200, Marcus Boerger wrote:

Hello Jani, we should change LEX/YACC/RE2C in gcov mode to omit those lines marcus Thursday, October 20, 2005, 9:25:53 PM, you wrote:

     You can get rid of the "bogus" #line lines with using
     the genfiles script. Don't even dream about getting rid
     of them in the files by using the -L or -l options for flex/bison!!

     --Jani

On Thu, 20 Oct 2005, Nuno Lopes wrote:

john Wed Oct 19 20:18:26 2005 EDT Added files: /php-src Makefile.gcov gen_php_cov Modified files: /php-src NEWS configure.in Log: Implementing C-level Code coverage (--enable-gcov). o Requires LTP 1.4+ and libgcov

Hello, As I've told you before, I had already tested your patch. It has the problem in the parsers that have bogus #line directives (but thats another story). The other problem is that you don't handle files with the same name. BTW, I have asked a server from my university for PHP automated testing. They told me that they would give me one, but I haven't received the code yet (waiting..). I have already done a little cron job to automates the things. The strongest point in the little programs I've developed is a patch for run-tests.php to let them run with valgrind, to check for mem leaks in each test. It is slow, but works very well. This particular patch I would like to see in CVS (see the phpt_diff_w.txt file in my public cvs) the stuff is at: http://mega.ist.utl.pt/~ncpl/cvs/viewcvs.cgi/phpqa/ Nuno