Re: [gnet] free heap block

[Andrewlanoix aol com]

The problems with the glib iochannels on windows is not so simple to describe as completely broken, for they are not. What Tor and I are saying is more nuanced -- that is the UNIX APIs allow things that we can’t support on windows. 

Owen responded: 
“My original conception was that GMainLoop would be significantly different on different operating systems, though GPollFD * crept into more of the API (g_main_context_query()) for
GTK+-2.x.”

Which basically says we should have developed a new set of low level APIs rather than try to support the UNIX ones. The new Windows specific APIs would only allow a subset of the things programmers have been accustomed to doing on UNIX.

So, if you are using the GNet API, you should be just fine since that only uses of subset of operations that are known to work and well tested on windows. The only open “true” bug on windows is that the main loop it relies on a wait function that has a limit on how many objects it can wait on. This has been in the code for years, and I have not heard any complaints. There are also some 64 bit problems I am sure in there somewhere.

Regarding GNet itself, it is like Glib, who also has lots of windows specific code which means that when new features are added it takes me months to add these one windows. GNet on windows is fully up-to-data as of my last check-ins which added IPv6 support on windows. This was painful to write given Microsoft’s documentation on the subject which is either horrible, no existent and in some cases wrong.

Andrew