I use glib’s testunit in visual studio 2002.But when compling the code, i find warnings.

The code:
…
g_assert_cmpint(3, ==, 2);
…

The warnings:
warning C4244: conversion from ‘__int64′ to ‘long double’, possible loss of data

In glib’s source code(gtestutils.h):
#define g_assert_cmpint(n1, cmp, n2) \
do { gint64 __n1 = (n1), __n2 = (n2); \
if (__n1 cmp __n2) ; else \
_assertion_message_cmpnum (G_LOG_DOMAIN, __FILE__, __LINE__, G_STRFUNC, \
#n1 ” ” #cmp ” ” #n2, __n1, #cmp, __n2, ‘i’); } while (0)

May be correct:
#define g_assert_cmpint(n1, cmp, n2) \
do { gint64 __n1 = (n1), __n2 = (n2); \
if (__n1 cmp __n2) ; else \
_assertion_message_cmpnum (G_LOG_DOMAIN, __FILE__, __LINE__, G_STRFUNC, \
#n1 ” ” #cmp ” ” #n2, (long double)__n1, #cmp, (long double)__n2, ‘i’); } while (0)

I use glib’s main event loop and iochannel, but why glib’s main event loop and iochannel don’t support epoll/kqueue/(/dev/poll)?

The poll/select func’s efficiency is very poor.

The libevent(http://monkey.org/~provos/libevent/) use epoll/kqueue. Why glib not?

thanks

May be you should consult the cunit’s code (http://cunit.sourceforge.net/).

2)I use glib’s Testing, but documentation is missing. There was no guide for how to set up your own project to use the glib testing framework or higher level overview of setup/teardown, available asserts, running, parameters, etc.

From experience I know that hard facts, code coverage percentage for example, encourages developers (or testers) to write complete test cases (feature wise) and eliminate unneeded or dead code.