[gpm]Re: new release: gpm-1.19.5

[Alessandro Rubini]

>> * After running ./configure in the "make" it re-runs autoconf and configure.
>>         autoconf may be not installed in the user's computer.
> 
> hmm...maybe it's just too late in the night, but
> on my system I ./configure was not rerun by make.

Hmm... on a fresh untar it does. Unfortunately you need to check the
tar before uploading. It's one of the boring tasks in being a maintainer.

This is what it does for me:

   [...]
   creating src/Makefile
   creating src/headers/config.h
   rudo% make
   autoconf
   if [ -f config.status ]; then /bin/sh config.status --recheck; \
   else ./configure; fi
   running /bin/sh ./configure  --no-create --no-recursion
   loading cache ./config.cache
   [...]

Using "make -d" on a fresh untar:

      Prerequisite `configure.in' is newer than target `configure'.
      Prerequisite `aclocal.m4' is older than target `configure'.
     Must remake target `configure'.

-rwx--x--x    1 rubini   staff       65075 Sep 20 23:02 configure*
-rw-------    1 rubini   staff        3063 Sep 20 23:27 configure.in

See, you didn't run autoconf before tarring. This is the difference
from the original autoconf and the regenerated one. Maybe you miss
a good "make distrib" rule. It was there (note that it used the
MANIFEST file, which is no longer valid but still there and unused).

> I wanted to seperate 'real' sources from other stuff, so I moved it
> to contrib. I did not mean to move it outside gpm,

Yes. But the src rules should not make t-mouse.elc
 
>> * The various "ex_tex" etc stuff is bugged, comparison always fails
>>         (no $$ in Makefile)
> 
> fixed & removed.
> I choosed another method:

maybe you just needed to have $$ex_tex instead of $ex_tex (see make docs).

>> Same for check_devfs and set_devfs, should be static into     liblow.c
> 
> How would you fix that most elegant ?
> #define lib_check_devfs static check_devfs  ?

Oh, didn't see that they were used in two places. Mabe you can make
them static inline in an header (gpmInt.h).

> LOBJ = $(LSRC:.c=.o) debuglog.o devfs.o @CURSES_OBJS@
> which is exactly what you wrote above.
> I am confused.

Ok, discard this (no time to check now).

>> * This in gpm.c:
>>                    chmod(GPM_NODE_CTL,0700);
>>         allows only root to connect to the server
> 
> I included this from a security patch. There it's told that this is used for
> security reasons. (gpm-1.19.1-secenhance.patch)

What security is it if the tool doesn't work?
Please check the source of security patches and the reason. What
secudity bug is this fixing?

> and chown it to mouse.
> If group mouse does not exist, use 0777,

Hmm... I'm not sure, really. Anyone should connect, and them gpm must
grant or deny permission. Do you really want mc, lynx and all gpm clients
to be group mouse, with suid? I think not.

>> * defines.h replicates a lot of stuff from gpm.h
> 
> I know. I partly wanted to cleanup the source.
> At the end there will be no doubled code anymore anywhere.

*please* remember that gpm.h is exported. You can't split them.
And gpm-defines.h etc is really ugly, not clean at all.
 
>> >    o new daemon name
>> Sorry?  Didn't find it.
> 
> Like Eastern in Germany I don't want to show you where the egg
> is. You will find it very soon ;)

Note that calling it gpmd or whatever is an incompatibility. People
is doing "killall -WINCH gpm" or such. Please don't add incompatibilities
unless needed, and when they are there, please mark them in huge letters
in the documentation and README files.

> maintaining a programs requires people finding bugs :)

:)

good work
/alessandro