Re: mc-4.5.99a

[Pavel Roskin <proski gnu org>]

Hi, Thomas!

> mine (I didn't see the contact address in the readme, so I looked in the
> changelog to see who would be a good contact).

Thanks for the patch and for the explanation.  Fortunately, I described
your changes close enough, and the complete explanation is now in the list
archives.

> 	a) the missing $ncurses_version caused resizeterm to not be found.

Clearly my bug.  Fixed.

Actually, the whole scheme for ncurses detection should be simplified.  
Currently MC cannot find ncurses on QNX Neutrino 6.1.0 because it mimics
old curses too closely.

> 	b) checking a report that mc didn't use the whole screen when setting
> 	   xterm font to "Unreadable", I noticed that mc died when I made the
> 	   screen very large.  That was a buffer-overflow problem when the
> 	   screen width was wider than 1024. (screen.c)

Thank you for finding it.  Having dynamic buffers just to survive this
"torture" would be a waste of resources, but boundary check is a good
idea.

> 	c) while debugging for the former bug, I tried using gdb with
> 	   ElectricFence.  But that didn't work.  The problem was that
> 	   mc's sigwinch handler was calling resizeterm, which performs
> 	   malloc's, etc., which essentially hung the process.  That was
> 	   probably leftover from my first patch in 1997 or 1998.  So
> 	   I ifdef'd out the (unnecessary) call.  (layout.c)

Thanks!

> 	d) before that ("c" took some work ;-), I found that mc was treating
> 	   the resize event as a SIGTSTP.  That was fixed by the ifdef in
> 	   key.c

I could not find any references to SIGTSTP in key.c.  My understanding is 
that your patch makes MC ignore KEY_RESIZE.  I wonder if the screen can 
actually be resized at this time.

> no problem.  When is the next version to be completed?  (So I know to come
> back and fix the mouse, if needed, before then).

I'm currently considering end of March for the 4.6 release.  
Unfortunately, my job in AT&T Labs takes too much of my time and resources
(yet it's a great place to work), but I hope to be able to work on MC
anyways.

VFS needs to be made more reliable (many thanks to Andrew Samoilov for 
doing so much not-trivial work on it).  MC should use color with S-Lang 
not only on a few "known good" terminals.  I want to scan the list for 
remaining unapplied patches.  Time permitting, MC should compile on Win32 
with MSVC and MinGW.  Also the regex problems on Cygwin should be sorted 
out.

Last week before the release will be dedicated to portability testing,  
This time it should be easier, because Automake is used consistently in 
all directories and there is no GNOME code.

-- 
Regards,
Pavel Roskin