Re: PCRE regex result is different between Linux & Windows.

Jochem Maas a écrit :

[I precede you, sorry for language mistakes...]

php or english? :-)

ohhh... sh.. ! I think I speak PHP better than english (silly, not ?).

okay, are you using the same PHP version on both machines? anything in the php.ini's that differs?

The same, not possible (Windows/Linux). For php.ini, quite the same (some directories are different). Under Windows (PHP 5.1.6 on 2k SP4 / 5.2.4 on XP SP2, the officials). Under Linux (Ubuntu 8.04) 5.2.4-2ubuntu5.3.

are you possibly looking at an input/file character-set encoding related issue? (i.e. encoding is different between the two servers)?

All PHP source is written in UTF-8. I take the HTML code and convert it to UTF-8 using iconv() / mbstring...

can you post a short complete script to see if others can reproduce the error?

See the following link for the bogus test (Must match : windows = 90, linux = 54): http://pastebin.com/m1c43cc10 The same results are given when : - comments are removed - with 'm' or 's' PCRE options - recursion is removed (multiple parses in while statement (matches for each pass : 55, 26, 5, 2)) This snippet is used in a part of code which goal is to convert HTML from Word 2003 to valid XHTML. But that is not the subject... For the PCRE version, I really can not tell you which one I use... Where can I see that ? So, It may be a bug ? Too bad...

have you tried to use the Tidy extension to clean up the input string?, it has alsorts of wonderful settings for making (x)HTML nice an shiny.

You think I have already tried it. ;-) Tidy is too agressive for parsing HTML from MS Office... Hope it will work.... :-/ -- Julien