MJPEG Tools Svn

Fix compiler warning on win32

Fix compiler warning - Process order ambiguous

Fixed the "read_header functions don't clear previous info first" problem:
o Two new public functions: y4m_clear_stream_info(), y4m_clear_frame_info().
o y4m_read_*_header() call these functions (respectively) at beginning.
o y4m_init_*_info() use these functions to set/init the default (non-)values.
Also:
o 'framelength' field in stream_info properly initialized.
o Moved y4m_fi_xtags() around to beautify source code.

pop up to 0.9.5 for next release

This commit was manufactured by cvs2svn to create tag 'driver-zoran-0_9_4'.

Oops

Add ChangeLog entry for all this

Update module use counting to whatever is used in 2.5.x. This is done quite evilly, I'm using an #if .. #else .. #endif everywhere. I'm doing this on purpose, because apart from the actual deprecation of MOD_INC/DEC_USE_COUNT in 2.5.x, there's also a lot of other cruft removed/entered that directly has to do with this, such as the removal of ->inc/dec_use in i2c_driver (i2c clients), and the introduction of a new struct member, ->owner. In most cases, I'm just substituting MOD_INC/DEC_USE_COUNT with try_module_get/module_put. This isn't actually tested, I just know it works. If it doesn't... Erm... Blame Hussein!

Fixed missed change from int to iterator that gcc-2.95.x didn't
complain about.

We don't actually need ISR for anything, so why do we continue using it in such a hackish way? Out, out