Re: Re: Question about emalloc

Ok, according to my OReilly "Programming PHP" book by Rasmus Lerdorf, the emalloc and efree functions work exactly the same as malloc and free. It says: ----- if you emalloc() something and forget to efree() it, PHP prints a leak warning like this if you are running in debug mode (enabled by compiling PHP with the --enable-debug switch): foo.c(123) : Freeing 0x0821E5FC (20 bytes), script=foo.php Last leak reported 1 time ----- so the docs do not agree... Im more apt to believe the book I have, since it goes on like this for a few pages about tracking down memory leaks... plus Rasmus wrote it ;) The reason why in your code there is no 'efree' is because of the '0' flag passed to 'RETURN_STRINGL'. If that was a '1', they would have to efree it. -- Brian 'Bex' Huff bex@stellent.com Phone: 952-903-2023 Fax: 952-829-5424 Mikey wrote: >>Im pretty sure you still need to 'efree' what you 'emalloc'. Every time >>I forgot to 'efree' something, I would get a bunch of error messages >>about memory leaks. The errors were nice and verbose, tho... as long as >>I was running the debug version of the dlls. > > > Which is what I would have assumed, but for what I read in the manual. > > >>So either the documentation is wrong, and you have to efree >>everything... or else the error messages about memory leaks that I keep >>getting can be ignored... >> >>I'd be curious to know the answer as well... > > > Well, I decided to see how the big boys dealt with strings, and had a root > around ext/standard. The following function is from string.c: > > PHP_FUNCTION(bin2hex) > { > zval **data; > char *result; > size_t newlen; > > if (ZEND_NUM_ARGS() != 1 || > zend_get_parameters_ex(1, &data) == FAILURE) { > WRONG_PARAM_COUNT; > } > convert_to_string_ex(data); > > result = php_bin2hex(Z_STRVAL_PP(data), Z_STRLEN_PP(data), &newlen); > > if (!result) { > RETURN_FALSE; > } > > RETURN_STRINGL(result, newlen, 0); > } > > As you can see, result is emalloc()'d, but not efree()'d at the end - looks > like the docs are right. I have since tried using this approach in a test > function and do not get any errors when I compile - what do you use for your > build? > > regards, > > Mikey > **All Electronic Mail sent from the Stellent, Inc Electronic Communication Network is scanned by Antigen 7.0.**