Created
May 17, 2010 01:26
-
-
Save tmpvar/403301 to your computer and use it in GitHub Desktop.
conversation about libeio and win32 compatibility with the man himself
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
<tmpvar> hello | |
<schmorp> hi | |
<tmpvar> hey, have you heard of any success with getting libeio built/running on windows? | |
<schmorp> no | |
<schmorp> it's not supposed to work, really | |
<schmorp> (windows doesn't support 99% or so of the required api) | |
<tmpvar> hrm | |
<tmpvar> yeah, good ol' windows. | |
* _ry ([email protected]) has joined #schmorp | |
<_ry> schmorp: what's wrong with libeio on windows? | |
<schmorp> wrong? | |
<schmorp> wlel, i guess, it works fine with e.g. cygwin | |
<_ry> <schmorp> (windows doesn't support 99% or so of the required api) | |
<schmorp> (there actually is workaround code for bugs in cygwin) | |
<schmorp> i assumed he meant native win32 | |
<_ry> yeah i want to use it with mingw | |
<_ry> (i know very little about windows) | |
<schmorp> then it's true | |
<_ry> it seems libeio doesn't do anything that's very magical | |
<_ry> except draining the queues at fork | |
<schmorp> well, except call dozens of functions thta don'T exist under windows | |
<schmorp> windows' doesn't even have the concept of a read or write to handle function | |
<schmorp> i have no clue how much mingw hacks around behind the scenes, of course | |
<_ry> you mean read(2), write(2) ? really? | |
<tmpvar> mingw doesnt really do much, its more of an effort to port unix binaries to run on windows | |
<schmorp> _ry: well, which one? | |
<_ry> eiher, both? | |
<_ry> i'm just not sure what you mean. that some of the eio wrappers doesn't exist on windows? | |
<schmorp> windows doesn't even have a read() function | |
<schmorp> well, none exist on windows | |
<_ry> wow.. how is that possible? | |
<schmorp> some are probably provided by wingw | |
<schmorp> mingw | |
<schmorp> but they will unliekly be working with threadds _and_ all sorts of file descriptors | |
<schmorp> windows itself doesn't have a generic I/O model | |
<schmorp> i.e. there isno equivalent to read() | |
<schmorp> WriteFile only really works on files, etc. etc. | |
<tmpvar> there is a read() on windows, but its for fd's only. socket() returns a SOCKET which is not an fd, although you can convert it to one using some low level api call | |
<schmorp> tmpvar: no, there is no reaD() on windows | |
<schmorp> tmpvar: you probably mean the libc of your compiler | |
<schmorp> which is different for each compiler... | |
<schmorp> cygwin has, of course, a read() | |
<tmpvar> sure, we'll be using libc via msys w/ gcc | |
<schmorp> so the quetsion is on what kind of "files" read and wirte even works | |
<schmorp> under posix, they work on everything | |
<schmorp> and if even read and write are missing | |
<schmorp> then sttavfs/unlink etc. are missing, too | |
<schmorp> windows, again, doesn't even have the very concept of unlinking files | |
<schmorp> and then you need pthreads | |
<schmorp> (but there is a workable pthreads library available) | |
<_ry> http://msdn.microsoft.com/en-us/library/1c3tczd6(v=VS.80).aspx | |
<schmorp> yes, but that doesn't unlink | |
<schmorp> it just tries to delete the file | |
<schmorp> besides, that one is called _unlink, not unlink... | |
<schmorp> and it's also not part of windows | |
<schmorp> as is clearly documented on that page | |
<_ry> i don't know. seems like getting some rudimentary support on windows is possible | |
<schmorp> sure | |
<tmpvar> http://msdn.microsoft.com/en-us/library/aa298551(v=VS.60).aspx -- _read | |
<schmorp> tmpvar: please do me the courtesy of not wasting my time, ok? | |
<schmorp> tmpvar: read the page you gave yourself | |
<schmorp> tmpvar: it clearly documents this as not aprt of windows | |
<_ry> schmorp: :) | |
<_ry> schmorp: does ev_async work on windows? | |
<schmorp> grrr | |
<schmorp> in theory | |
<schmorp> the problem is that it relies on nonstandard extensions | |
<schmorp> so in practise, it doesn'T work | |
<schmorp> because read/write etc. doesn'T work on sockets | |
<schmorp> (they work on pipes...) | |
<_ry> ev_async only uses a pipe, right? | |
<schmorp> cannot on windows | |
<schmorp> windows doesn't even have the concept of unix-style pipes | |
<_ry> so i odn't know how mingw works - but i imagine it has glibc? | |
<schmorp> wow | |
<schmorp> of course not | |
<_ry> okay | |
<_ry> i'm optimisic though - even if we have to #ifdef in a new ev_async | |
<schmorp> ev_async doesn'T even exist in libeio | |
<schmorp> i think you swithed topics | |
<_ry> yeah but i use both | |
<schmorp> libev works under window, with some love | |
<schmorp> except that a number of people mentioned thta read/write doesn't work for them | |
<schmorp> which makes sense, as windows doesn't have those | |
<schmorp> so everybodies read/write comes from a different codebase that hacks around windows | |
<schmorp> while it could reqaosnably be done by porting to a single such environment | |
<schmorp> it's quite a diffeent thing to make ti work with all such | |
<schmorp> as such, libeio is fully supported and works on windows | |
<schmorp> as long as you use cygwin | |
<tmpvar> thanks for your help | |
<schmorp> good luck |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment