include/.gitignore should continue to ignore old files in case
someone checks out an older version, builds, then checks out the
current version without running 'make spotless' first.
sys/unix/Makefile.utl: tiletxt.o depends on tilemap.c in addition
to tiletxt.c.
observed: parallel build attempts of makedefs that trampled over
one another.
attempted workaround: Add a dependency as per Pat R's suggestion.
observed: Concurrent header file movement collisions were sometimes
causing file busy errors and build failures.
workaround: Eliminate tile.h header file movement from the
Makefile build so that the collisions won't occur with that
particular file. Leave the header file tile.h in win/share as it
is in the distribution and just adjust the include path in the
rule for the specific files that use it.
observed: tiletxt.c created on-the-fly from Makefile echo statements
sometimes resulted in garbled and duplicate content in it when
parallel makes were involved, and that caused a build failure.
workaround: Instead of creating a tiletxt.c on-the-fly via echo
statements in the Makefile, simplify things and use that
same #include "tilemap.c" approach but make it an actual file
in the distribution. That makes it available for other platforms
too.
Place built libraries for Lua and pdcurses into lib instead of the
more transient src/o subfolder.
Remove a kludge involving sys/windows/stub-pdcscrn.c.
Don't link pdcurses into NetHackW.exe (required a couple of stubs
since NetHack.exe and NetHackW.exe currently share object files
under the visual studio nmake build.
(Note: This may require a couple of follow-on minor modifications
to the mingw build. If so, the CI will flag that for us after this
commit)
Requested by k21971 (hardfought admin): classify the gameover event
as something other than achievement so that live-logging can exclude
it in order to use the xlogfile end-of-game entry instead.
It had been classified as an achievement because that was the only
category being treated as 'major' so written to final dumplog. Give
is a new classification 'dump' which is distinct from achievement
and intended to explicitly request that an event go into dumplog.
The gameover event is the only one in that category, at least for now.
Add a bunch of other classifications besides achievement and dump to
be treated as 'major' for dumplog. The new list is
wish, achieve, umonst (death of unique monster), divinegift,
lifesave (via amulet, not explore-/wizard-mode decline to die),
artifact, genocide, dump
and may still need further tuning. Currently excluded are
conduct (first violation of any conduct), killedpet, alignment
(changed, either via helmet or conversion), minorac (minor
achievement such as level gain or entering the mines), and spoiler
(currently only applies to finding Mines' End luckstone which is
also 'achieve' so will be in dumplog).
This doesn't remove the reference to unimplemented LLC_TURNS that is
mentioned in the template sysconf.
Closes#688
From copperwater: a recently added use of <test> ? <if> : <else>
had a ptrdiff_t (signed) expression for <if> and a size_t (unsigned)
expression for <else> which triggered a sign-compare warning when
the two expressions are implicitly converted into the same type.
Use casts to convert both expressions to long rather that convert
the size_t half to ptrdiff_t or vice versa. The final result gets
cast to int already.
Fixes#689
src/moc.qt[56] existed briefly but got replaced by src/Qt[56].h-t
(controlled by hints/include/compiler.370). The old names were
deliberately left in src/.gitignore and sys/unix/Makefile.src's
'clean' target to give some time for anyone who had generated them
to run 'make clean' to get rid of them. That time is now up. It's
only been about three weeks but if I wait any longer I'll probably
forget.
For the !SYSCF configuration, the command line processing still checks
for a value for maximum number of simultaneous players. The recent
revisions would have accepted a negative value. I don't know whether
anything interesting would have happened if someone did that.
Move a bunch of stuff out of main() into new early_options(): '-dpath'
playground directory handling, '-s ...' show scores instead of playing,
and the 'argcheck()' options: --version, --showpaths, --dumpenums,
and --debug (not to be confused with -D). Also introduce
| --nethackrc=filename
| --no-nethackrc
to control RC file without using NETHACKOPTIONS so that that is still
available for setting other options. They can start with either one
or two dashes. --no-nethackrc is just --nethackrc=/dev/null under the
hood. '-dpath' can now be '--directory=path' or '--directory path'
but the old syntax should still work. '-s ...' can be '--scores ...'.
Basic call sequence in unixmain relating to options is now
|main() {
| early_options(argc, argv[]);
| initoptions(); /* process sysconf, .nethackrc, NETHACKOPTIONS */
| process_options(possibly_modified_argc, possibly_modified_argv[]);
|}
Options processed by early_options() that don't terminate the program
are moved to the end of argv[], with argc reduced accordingly. Then
process_options() only sees the ones that early_options() declines to
handle.
Most early options were using plain exit() instead of nh_terminate()
so not performing any nethack-specific cleanup. However, since they
run before the game starts, there wasn't much cleanup being overlooked.
chdirx() takes a boolean as second argument but all its callers were
passing int (with value of 1 or 0, so it still worked after being
implicitly fixed by prototype). Change them to pass TRUE or FALSE.
argcheck() was refusing (argc,argv[]) with count of 1 but then it was
checking 0..N-1 rather than 1..N-1, so it tested whether argv[0] was
an argument instead of skipping that as the program name. Change to
allow count of 1 with modified argv that has an option name in argv[0].
That happens to fit well with how early_options() wanted to use it.
The checking for command line flags --version, --showpath, and one
or two others was inside #if CHDIR. I don't know whether anyone
ever disables that configuration option, but it shouldn't control
whether those flags are handled.
I wanted to be able to specify -windowtype:foo on the command line so
that I didn't have to use "NETHACKOPTIONS='windowtype:foo' nethack"
and it turned out that such an option already exists, as "-wfoo".
I either never knew about that or had completely forgotten it. Anyway,
this makes specifying windowtype be more versatile.
"-wX11" still works; now "-w X11", "-windowtype=X11", "-windowtype:X11"
work too, with "--" variations of the latter too also supported. The
long name can be truncated to any leading substring of "windowtype",
although it has to be at least "wi" for "--"; "--w" is rejected.
Also, any errors reported while processing the command line are
treated like config file processing errors rather than just delivered
with raw_printf(). On tty at least, they used to vanish when the
screen cleared to start the game, with no chance to read them. Here's
an example from after this change. It sets windowtype to tty and then
overrides that with X11.
|% ./nethack --w:Qt --win tty -wX11 -windowtype
|
|
| * Unknown option: --w:Qt.
| * Window type [nothing] not recognized. Choices are: tty, curses, X11, Qt.
|
|2 errors on command line.
|
|
|Hit return to continue:
This should probably be better integrated with argcheck() or vice
versa but the only change to that was a couple of formatting bits.
Anything that already worked should continue to work just the same,
aside from the improvement to the error feedback.
Turning on -Wformat-noliteral for Mac triggered a new warning.
Blindly suppressing the warning would have silenced it but would
also have left a real bug in place. The former format was passing
a string argument to %d format.
This converts the format to a literal with an additional argument
for the non-literal part. It compiles cleanly but I don't know how
to test it, let alone force an error for it to report.
Log game events, such as entering a new dungeon level, breaking
a conduct, or killing a unique monster, in a new "Major events"
chronicle. The entries record the turn when the event happened.
The log can be viewed with #chronicle -command, and the entries
also show up in the end-of-game dump, if that is available.
This feature is on by default, but can be disabled by
defining NO_CHRONICLE compile-time option.
This also contains "live logging", writing the events as they
happen into a single livelog-file. This is mostly useful for
public servers. The livelog is off by default, and must be
compiled in with LIVELOG, and then turned on in sysconf.
Mostly this a version of livelogging from the Hardfought server,
with some changes.
Report stated a -Wformat-nonliteral at line 612,
and a -Wformat-security at line 614
I was only seeing the latter, so I added the former to the
flags in sys/unix/hints/include/compiler.370. Some compiler
versions have that warning on by default internally and others
don't. If the format string isn't a string literal, there's no
inteference with printf argument checking because that only
operates on string literals.
NetHack 3.7 stores timestamp information, as well as github
commit hashs information if available, internally by compiling
date.c. It is important to ensure that date.c is always
recompiled after any other NetHack source files are compiled.
As usual for the visual studio nmake Makefile.msc, steal the generated
dependencies from the tail of sys/unix/Makefile.src, and adjust the text
to suit the Windows build.
Rename the recently added timestamp file used to throw away old qt
'moc' files from moc.qt5 or moc.qt6 to Qt5.h-t or Qt6.h-t and use
that to also throw away old qt_*.o when switching from Qt 5 to Qt 6
or vice versa. Temporarily the old names remain in Makefile.src's
'clean' target and in src/.gitignore but those will be removed soon.
Update 'make depend' to add the timestamp file to qt_*.o dependencies.
Have it generate rules to build qt_*.moc from ../win/Qt/qt_*.h instead
of using a template rule in hints/include/compiler.370. So building
the Qt interface doesn't require use of that hints file anymore and
someone reading Makefile.src won't have to know about it, but using
those hints will make their life easier.
Simplify the Qt timestamp handling portion of compiler.370. Only one
extra rule gets added when creating src/Makefile.
src/Makefile generated from sys/unix/Makefile.src that's been rebuilt
with 'make depend' got broken by uncommenting '#include "amiconf.h"'
in global.h. That file isn't in include/ but every object file now
depended on it and make didn't know what to do about that. Have
depend.awk treat it as a special case so that no object files depend
on it. That means that actually modifying it won't trigger a rebuild;
anyone fiddling with that will have to always do 'make clean' or
'touch config.h-t' after changing it. The alternative is to move it
from outdated/include/ back to include/.
In depend.awk, recent gawk complained that "\." wasn't a defined
escape sequence in regular expressions so it would be treated as ".".
That's exactly what is intended but change it to "[.]" to avoid the
warning. Similarly for one instance each of "\#" and '\"'. I also
tried changing "\/" to "[/]" even though that is a defined sequence
and doesn't trigger any warning. gawk accepted it but the awk that
comes with OSX choked on it so I changed it back to "\/".
Get rid sys/share/cpp[123].shr, the pre-ANSI C preprocessor that was
included in the source distribution for use on systems with ancient
C compilers whose preprocessor that couldn't cope with nethack's large
number of macros.
