1.44
Modern g++ seems to dislike *printf without a constant format string if there
are no other aguuments and was causing compiles to fail. Fixed warning
and patch from Tobias Quathamer.
Manpage spelling fix from Tobias Quathamer.
John Heidemann pointed out that plot doesn't accept modifiers to strings
(though he didn't phrase it that way). The short form is that
plot bullet at x,y
wasn't working. I've adjusted plot to make the string modifiers into
active keywords, which fixes the problem.
Someone named Y T pointed out that there was no mod operator (%). There
is now.
This marks the transition from my home brewed Makefiles to automake-generated
makefiles. A couple largely unused options went away, but with any luck
we still build everywhere.
Robert Daasch submitted a patch to include floor and ceil fuctions.
That's been incorporated. (His patch also included adding a mod
operator, but I'd already done that.)
Added minimal support for dates, by adding versions of strftime and
strptime that return and accept seconds since the unix epoch.
1.43
Bruce Lilly pointed out that g++ 4.3.1 was emitting all sorts of dire
warnings about the hash_map data structure and associated support being
pulled from g++ in the future. SuSe seems to ship with this and there's
no point scaring people (or letting a time bomb tick). Grap now detects
and uses unordered_map if it's present on your system, which stops the
wailing and gnashing of teeth from g++ 4.3.1 . That g++ version also
requires a compiler option to support that include file, which we also
autodetect and use. (This is essentially autodetecting the flags described in
the changes to grap 1.41 below, along with some code changes to make that
actually function correctly.)
A fellow named Fergus had a Cygwin compilation problem as well. His
system apparently had rand but not random (a very rare configuration
these days). This led me to find a corner of the autoconf code that I
apparently hadn't checked sufficiently. If rand was found and a
declaration made in the system files, the grap config stuff failed to
note the declaration and made its own incompatible one. This should be
gone now.
Fergus's system also seemed to be lacking snprintf - which again is very
odd. I hadn't reflected the change to grap sprintf in version 1.23 into
the code that emulates snprintf on systems without it. That code has
been added.
Changed the examples to include a brief tutorial on string matching in
grap. Suggested by John Heidemann.
1.42
There's been a long standing bug with how different versions of pic
interpret the "line from (x,y) then down 5" construct. There was once a
bug report at
http://lists.gnu.org/archive/html/groff/2008-03/msg00003.html
about this. As of version 1.42 grap no longer outputs this pic construct.
Dan Lasley pointed out that the key was incomplete for the bar graph example.
His fix to restore the key is included.
John Heidemann points out that bars in coordinate systems other than the
default just didn't work. This was a bug in my yacc grammer - for
heaven's sake. It's corrected to match the manual page.
For loops were strange in that
.G1
for xx from 15 to 13 do { print "YYY" }
print "ZZZZ"
.G2
would loop infinitely. xx never passes 13. Added a test to make that a
null loop as well as making
.G1
for xx from 13 to 13 do { print "YYY" }
print "ZZZZ"
.G2
print one "YYY". You can still confuse loops, using strange by clauses
- for example "by 0" or "by * -1" but common cases should be covered.
Spotted by a fellow named Yuval.
Found a lurking initializer bug while fixing the for loop thing.
I added references to standard plot strings in a couple places to make
them somewhat easier to find.
c029bdd1d9