Re: com php-src: fix default args for --with-pcre-valgrind: ext/pcre/config0.m4

On 25/07/17 00:31, Christopher Jones wrote: > > > On 20/7/17 10:31 am, Anatol Belski wrote: >> Hi Chris, >> >>> -----Original Message----- >>> From: Christopher Jones [mailto:christopher.jones@oracle.com] >>> Sent: Thursday, July 20, 2017 2:11 AM >>> To: Anatol Belski <ab@php.net>; php-cvs@lists.php.net; Tianfang Yang >>> <tianfang.yang@oracle.com> >>> Subject: Re: [PHP-CVS] com php-src: fix default args for >>> --with-pcre-valgrind: >>> ext/pcre/config0.m4 >>> >>> Hi Anatol, >>> >>> 7.2 & 7.3 builds now fails for me on macOS and Oracle Linux 7 with: >>> >>>     checking for Valgrind headers location... configure: error: Could >>> not find >>> valgrind/memcheck.h >>>     Configure failed >>> >>> Do we really want users to have to explicitly disable PCRE valgrind >>> when they are >>> using --enable-debug? >>> >> It wasn't by default in the first version, later it was suggested by >> Nikita and for debug mode certainly makes sense, as for me. The reason >> behind this option is the upgrade of the bundled PCRE, which shows a >> lot of false positives without the integrated Valgrind support. Also I >> was basing on the fact the debug mode is actually developers only, >> normal use wouldn't be affected therefore. Why would users do debug >> builds? >> >> On a dev machine Valgrind would be anyway present. For developers it's >> best to have it by default, as PCRE usage is spread around the core >> here and there. Bug fixing also does normally involve valgrind checks. >> I wouldn't see the option itself as a big question, either way. If >> many people find the default enablement unsuitable, so the behavior >> can be reversed. >> >> Regards >> >> Anatol >> > I'm not totally convinced merging debug-level arguments makes sense (why > not always enable --enable-phpdbg-debug too?)  But there are more > important things to worry about, so I won't stress. > > BTW, './configure --help' prints the --with-pcre-valgrind=DIR help twice. > > Chris > Just stumbled over this one, too. I don't think it's very nice to suddenly let ./configure --enable-debug die when valgrind headders are not present, and I didn't request valgrind support in ext/pcre. Suddenly all of my travis builds failed, for no apparent reason, despite --disable-all I think it should only fail hard, if explicitly requested. -- Regards, Mike

Attachment: [application/pgp-signature] OpenPGP digital signature signature.asc