Hi, On Sun, 2008-08-03 at 22:15 +0200, Pierre Joye wrote: > Hi Derick, > > On Sun, Aug 3, 2008 at 6:34 PM, Derick Rethans <[EMAIL PROTECTED]> wrote: > > > you're not reading what I said. It does not make one single bit of sense > > that a short function gives problems while an enormously long function > > is fine. This does *NOT* make sense. If you have no idea either, then > > figure it out first before you add silly #pragmas to code. > > This code broke the VC6 build, that's a fact. A breakage must be > fixed, period. I figured out why and fixed it, end of the topic. Let > forget the insane amount of warnings in the date code... (top #1 in > PHP).
Correct PHP is supposed to work on "all" platforms therefore the code has to work on Windows, too. Now there are two questions Derick ask, which I see as valid questions: a) Why does it break with that function while at least another bigger one exists. Do we have to expect the other one breaks too the next time the scan() function is regenerated from re2c for instance? It feels a bit awkward that a small function is too big for the compiler while a way bigger one isn't. So finding the exact reason might help in the future. That's mainly curiosity. b) In general the build system is not only for putting the stuff in the compiler but also detecting compiler problems and working around them. Is it possible to do this work around in the build file? The benefit would be that it's less likely that a developer breaks it by accident and we avoid making our code harder to maintain by not cluttering it with compiler specific stuff (probably the Aix compiler has a problem with some other stuff, the Sun compiler with other stuff and yet another compiler somewhere else) Yes one workaround doesn't hurt but let's wait 5, 6, 7 years of further growth of the code base and new compilers with their issues to be supported... Therefore such changes should be, when possible, be placed in the build files. Oh and also see the discussion about a gcc issue we recently discussed on the list. So as a summary: The nicest code is worthless if we can't compile it on our platforms, but build fixes should be places where they hurt the less. johannes -- PHP Internals - PHP Runtime Development Mailing List To unsubscribe, visit: http://www.php.net/unsub.php
