-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On 25.01.2014 17:50, Evgeny Grin wrote: > > > 25.01.2014, 17:34, "LRN" <[email protected]>: >> This is why wrappers may still have to remain. Sprinkling >> W32-specific code all over MHD is not something grothoff will >> appreciate, i'm sure. Wrappers move W32-specific parts into >> separate source file, and look the actual MHD code look at least >> a bit sane. > > Code can be rewritten as platform-agnostic. Not much work is > required for it. Only two places with noticeable amount of win32 > code: VirtualAlloc instead of mmap and pipe/socketpair. That's > all. Actually, i meant the error-handling code. Imagine you've called accept(), and want to check the errno for possible errors. But WS2 version of accept() doesn't set errno, it sets LastError. You need to convert LastError to errno, otherwise your errno check won't work. Adding conversion function call after the function call might work in some (most, actually) cases, if it's hidden under a platform-specific macro that expands to nothing on non-W32 platforms, but using it might be error-prone (not doing this call will result in subtle runtime behaviour change that will be difficult to catch, and only affects W32).
mmap and socketpair are the last things i'd be concerned with. They are definitely going to take shape of separate functions, albeit with simplified implementation, where applicable. - -- O< ascii ribbon - stop html email! - www.asciiribbon.org -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.11 (MingW32) iQEcBAEBAgAGBQJS48SGAAoJEOs4Jb6SI2Cw1GEIAJVA/XHjb/ykgiGQVaRl+EYN HXWWo6yT4hb7vl4rKvzrJGuhs5j3F/gnXwUPxvVBrIFa8lvXRVLqxIyF6rQ2qg5p 2Nyo2ZzMO1jQI8BF1+aus6Ozk0XVC/SeyhcunjVJ6cgkGy1YCKO96V5rqB3h+dp5 AynxMCAlIkAwF327er4iSTMgJ0F3lgLZg9aJRbqYpCQUR/k1iQAhhOKYibkMH27G AN6WlFCdBXwswOmQQrPdWXvmnLmhbIOMmCACiG7qpkrYN2syo3X2xawVUbNU1aU8 ZtMXqz+qbKY2IBKX0e0jUZqldmhyj8PGACftriCGCgVUvQZqioj+fchChz4U97o= =PvYg -----END PGP SIGNATURE-----
