mc-4.6.1 release?

[Roland Illig]

Pavel Roskin wrote:
> On Thu, 21 Oct 2004, Roland Illig wrote:
> 
>> I would like mc-4.6.1 to be released soon. The broken pipe issue won't 
>> get away in the next few days, so there's no reason to wait for it. I 
>> have a still uncommitted patch which fixes the error message to be not 
>> as cryptic as before.
>>
>> http://www.schlechte-software.de/pub/mc-patches/mcview-sigpipe.patch
> 
> It's unfortunate because it will be perceived as a regression, even 
> though the code in 4.6.0 was incorrect.

I know. The problem is our usage of open_error_pipe() and 
close_error_pipe(), which is bad by design. (It will run into a deadlock 
if the child process writes more than a few kbytes to stderr; stderr 
writes directly on the screen in the "paste output of" editor command.) 
There will be an alternative soon. I am already working on it.

> OK, we need some testing before the release.  maint/mctest gives some 
> warnings on Fedora Core 2.  From test-2.err:
> 
> ../../vfs/mcserv.c: In function `mc_pam_auth':
> ../../vfs/mcserv.c:782: warning: assignment discards qualifiers from 
> pointer target type
> ../../vfs/mcserv.c:783: warning: assignment discards qualifiers from 
> pointer target type
> ../../src/cons.saver.c:52: warning: no previous prototype for 
> `send_contents'
> ../../src/cons.saver.c:105: warning: no previous prototype for `die'

I have fixed them. The changes were trivial. The warnings had been there 
because I don't HAVE_PAM and so I didn't notice them.

> We also should test FreeBSD and Cygwin.

I have just collected the 28 build logs of my test configurations on 
SPARC/SunOS-5.9. They all failed because INADDR_NONE is not #defined. I 
have committed patches for these, too and will rerun the testsuite.

Roland