Several small changes, like freeing pixmaps which won't be needed
later, reducing the default Imlib2 cache size to 0 bytes (we don't
really use it well), and an image path bugfix spotted by Teodor
The big change, of course, is the disappearance of libmej and the new
dependency on libast, which is now distributed separately.
SVN revision: 3793
I'm working on some deallocators now. The idea is that when Eterm
exits and memory debugging is on, several routines get called to free
the in-use memory (menus, font cache, etc.) that we still know about.
Anything left after that would be either unavoidable leaks (like
environment variables...read the putenv() man page sometime...sigh)
or genuine memory leaks that need fixing. I'm down to about 4.5K of
leftover malloc'd memory now. Making progress....
SVN revision: 3295
Okay, everything seems hunky-dorey now. If you have memory debugging
turned on, exiting Eterm will give a dump of the leftover allocated
memory including where it came from and how much there is. You'll
also get a listing of all the remaining Pixmap ID's and GC ID's along
with where they were created.
I also added some macros to the debugging stuff in libmej.h so that
if you pass the --without-debugging option to ./configure, it will
optimize out all the ASSERT and REQUIRE goop. This will make it
harder to trap bugs before they cause problems, so don't do it if you
want to help development, but if speed is critical to you, give it a
SVN revision: 3294
That should fix the crashes. Somehow I managed to temporarily forget
about pointer arithmetic. And somehow I thought trying to dereference
a pixmap ID would be a good thing. Sigh. I need sleep.
SVN revision: 3292
Massive reorganization/rewrite to libmej. It should now be 100%
independent of Eterm. There still may be some gremlins in the memory
debugging code, so don't use too high a number with --debug....
SVN revision: 3282
Color modifiers for images now work once again. Please report any
problems you find with them as soon as possible; I want to get 0.9.1
out the door in the fairly near future since the TODO list for 0.9.2
is already sizeable.
SVN revision: 3241
Two more patches from Marius Gedminas <email@example.com>. The first
one allows for customization of the message displayed when Eterm goes
into pause mode. There are actually two; one goes in the titlebar,
and the other is displayed in the text window.
His second patch makes Eterm's behavior a little smarter when it
resizes itself. It tries to figure out which quadrant of the screen
it's on and resizes in the most appropriate direction. (For example,
Ctrl-GreaterThan on an Eterm in the lower right corner will cause the
upper left corner of the Eterm to move; the lower right corner will
Thanks again to Marius for saving me time by sending patches. :-)
SVN revision: 2912
The multibyte font size patches caused weird behavior if your
multibyte font sizes didn't match your regular font sizes for those
who don't actually use the multibyte support. So I fixed that. I
also fixed the XA_TEXT problem for people using old/broken X.
SVN revision: 2904
Several people have asked for the spec file to be in the tarball so
that rpm -ta <tarball> works properly. I've put it in there, but I
don't recommend doing it that way. I use /usr/local as my prefix, and
there's a good chance I'll forget to change it before doing a dist
I also put some dummy crap in pixmap.h so that compiling without
Imlib2.h works again.
SVN revision: 2903
I think this --pause mechanism will be agreeable to everyone. You can
once again exit with any key (well, any key that has an actual
character/string associated with it by X...not Insert, Home, etc.),
but keys Eterm uses (like Shift-PgUp) will still work as expected.
SVN revision: 2902
Marius Gedminas <firstname.lastname@example.org> reported a couple of issues back
when he was on his bug-spotting binge that I didn't have time to fix
just then. Well, now I've fixed them. ~/.Eterm/user.cfg will now
be found if there isn't a theme-specific one; this allows you to have
a single user.cfg which specifies some options you want all your
Eterms to have.
Along those same lines, the action code now searches for duplicate
bindings and changes the existing one rather than adding a new one
to the end of the list. This allows bindings in user.cfg to override
those in theme.cfg (as they should). Also, bindings are added in
reverse order, so newer ones (like in user.cfg) take precedence over
older ones (like in theme.cfg) if there is a conflict (e.g., if your
theme.cfg binds "anymod button2" and user.cfg binds "ctrl button2,"
SVN revision: 2901
Fixed an old rxvt bug in the handling of "\e[0m". I *really* wish I
had time to rewrite that damned thing. Such a piece of crap. Thanks
again to Marius Gedminas <email@example.com> for pointing it out.
SVN revision: 2780
Fixed a seg fault when resizing to one row while on the secondary
screen. I have no clue how that bug crept in there all of a sudden,
but thanks to Marius Gedminas <firstname.lastname@example.org> for pointing it out.
SVN revision: 2772
Dammit! I did something really stupid. For anyone who doesn't know
already, calling X from a signal handler is a NO-NO. The *really*
stupid part is that I already KNEW that! *slaps self*
SVN revision: 2745
Okay, first off I need to thank Marius Gedminas <email@example.com>.
Not only did he point out several issues, he was even willing to send
a patch. :-)
This commit includes Marius' patch which keeps MappingNotify events
from overriding users' modifier settings. It also includes a fix to
a seg fault that he pointed out when menus were loaded without the
menu imageclass having been defined.
He also pointed out that --pause tends to sit and spin, taking up 100%
of the CPU time. Funny how E-Cpu never seemed to get too bent out of
shape over it, but he was indeed correct.
Thus I have fixed it, and while doing so, I have changed the way it
works. It no longer waits for a keypress per se. Actually, it just
ignores the fact that its child went away and keeps right on taking
X events. There are a few exceptions though. Either ESC or Ctrl-C
will exit a paused Eterm. Any other input that doesn't have special
meaning to Eterm will be ignored. (Shift-PgUp and Shift-PgDn still
work however, as do any action bindings you may have.)
SVN revision: 2739
My attempt to track down the strange X errors have revealed that a
pixmap given to me by Imlib2 is getting freed somehow behind my back.
Probably because this part of Imlib2 hadn't been tested before Eterm
was converted. =P
It doesn't seem to happen in XFree86 4.0; I'm wondering if XFree 4 is
smart enough to detect double-frees of old XID's and just ignore them?
Well, I'm going to have to add some debugging code to Imlib2 and see
if I can track down where it's freeing my pixmaps. But I have some
errands to run first, so I'm going to commit this for now. It
shouldn't actually change any functionality.
SVN revision: 2722
Okay, there are a few changes here. First off, I made multi-byte font
support the default now, as long as you have ISO 10646 fonts. In
order to do this, I made the default encoding type "Latin1" so as not
to interfere with 8-bit ISO 8859-1 characters. This means that if you
relied on the default multi-byte encoding method to be SJIS, you'll
need to update your theme files.
I also set it up so that Eterm will ignore SIGHUP, at least until I do
something with it (like reloading the theme or something).
I fixed the proportional font size algorithm. If there is more than
a 3-pixel variance between the minimum and maximum sizes for glyphs in
a proportional font, Eterm will set the size to 2 standard deviations
above the average width. This is so that they won't look so spread
out and ugly, but it still doesn't look perfect. Not much I can do on
that front...terminals must have fixed-width columns.
And then there's the biggie. I put in the ability to configure the
now-infamous font effects. I left a black drop shadow in as the
default, but you can now customize it via the --font-fx option or in
the config file using "font effects <stuff>" in the attributes
context. You can even use "fx" instead of "effects" for short.
So what goes in the <stuff> part? Well, you have several options.
To use a single-color outline, say "outline <color>". Likewise, a
single-color drop shadow is "shadow [corner] <color>"; "bottom_right"
is the default corner if you don't specify one. For a 3-D embossed
look, "emboss <dark_color> <light_color>". The opposite, a carved-
out look, can be had with "carved <dark_color> <light_color>". (Of
course, with those last two, the 3-D look will only work if you
choose the colors wisely.)
Those are all the shortcuts. The long way is to specify a series of
corner/color pairs, like "tl blue" for top-left blue, or
"bottom_right green". You can abbreviate using "tl," "tr," "bl," or
"br," or you can spell out "top_left," "top_right," "bottom_left," or
"bottom_right." If you omit a corner name, the first one defaults to
top-left, the second to top-right, and so on as listed above.
SVN revision: 2714
Added new shade/tint routines, including some done in MMX assembly, by
Willem Monsuwe <firstname.lastname@example.org>. Thanks Willem! These should prove
to be faster than the old stuff, probably by quite a bit.
SVN revision: 2595
Fixed a bug in the handling of resize events thanks to a tip from
Teodor Zlatanov <email@example.com>. While I was at it, I also fixed a
couple remaining issues with the popup menu system. You shouldn't
get any more cases where items remain selected after you've moved your
mouse off of the menu, or where a submenu would stay up even though
another menuitem is selected.
SVN revision: 2594
After getting sick of hearing people whining about the obscure error
messages that resulted from trying to build without pixmap support, I
hacked it until it at least built without it. I also added warnings
to configure.in so that people would know if that happened.
It's still not very happy running that way, which is something I need
to fix. But hopefully it will stop the whining.
Yeah, right. =P
SVN revision: 2590