| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
| |
There's really no reason not to show the line; in other options, having
the default always be the case can be annoying, but if you don't want
to look at the gold/turns display, you can just not.
|
|
|
|
|
|
| |
The amulet of conservation and cloaks of preservation are no longer
necessary without item destruction. Maxwell's patent armour now has
the resistance ego and grants rF+ and rC+.
|
| |
|
| |
|
|
|
|
|
|
| |
Unlike being run by hand during actual debugging, during stress testing you
almost always want to get the backtrace. This also helps solving recent
crop of mysterious pthread_join segfaults.
|
| |
|
|
|
|
| |
git would remove the directory otherwise.
|
|
|
|
| |
Although, it doesn't do anything stress tests wouldn't catch.
|
| |
|
|
|
|
|
|
|
| |
This means there's no dynamic loading of maps anymore.
Also, delete unused regression tests for the long gone old LOS code.
We had copies of both the old and current version.
|
|
|
|
|
|
|
|
|
| |
The setting is applied only on reload, every 1000 time turns. This means
that after a newarea shift you'll get only ordinary abyssal monsters, but
that's good enough.
Why Forest and newCrypt got such a huge weight is left as an exercise to
the reader.
|
| |
|
|
|
|
| |
Easier to predict when the time chunk will end.
|
|
|
|
|
|
|
|
|
| |
Restarting from such a save would reset the random seed, while the first run
had a random generator hold a different state.
This doesn't handle real level transition (ie, the initial banishment and
being thrown deeper into the Abyss). That's doable if we would inhibit the
commit somehow, but for now, I don't care much about those 5000 turns.
|
|
|
|
|
| |
Because, you know, being able to reproduce crashes without having to run a
random bot for days, might be of some use.
|
|
|
|
|
|
| |
After stat death has been removed, it stays all the time with negative stats,
even with living races. There's just that much stat drain in the newabyss,
especially int.
|
| |
|
| |
|
|
|
|
|
| |
There's nothing interesting that bad mutations bring to the testing, and
there's Xom already.
|
|
|
|
|
|
|
| |
It spends most of the time in stat death otherwise, with all the stat drain
abundant in the Abyss. This means monsters get to move when we should use
the CPU more for exploring the Abyss (I don't disable monsters here to get
a better test coverage).
|
|
|
|
|
| |
Newnewabyss has hardly any liquids, and it doesn't let us check fpr falling
into water.
|
| |
|
| |
|
|
|
|
|
| |
No point in stretching it, it should be enough for benchmarks, valgrind
or profiling, and you need an extended run to catch any randomness anyway.
|
|
|
|
|
|
|
|
|
|
| |
One includes holding '5' with monster spawns disabled; useful for
benchmarking Abyss morphs with no shifts other than timed ones.
The second one is the Abyss bot, limited to 1000 turns. It's about
twice as slow as the former, testing monster spawns as well, combat,
scrolling shifts. Unlike the bot, it doesn't use Xom, as that costs
yet another ~20% speed (we probably should optimize tension, too!).
|
| |
|
|
|
|
|
| |
This allows redirecting the results. Usually you'll want to put STDOUT to
/dev/null, too (perhaps we should to that always?).
|
|
|
|
|
|
|
|
|
|
| |
Otherwise, it'd unfairly charge the first run the cost of rebuilding.
We take the median of five so it wasn't that bad, though.
Also, load the db (including parts which were already ready) into the
system's page cache, and flush outstanding writes. We take user time
rather than wall clock so it shouldn't matter, but that's theory not
practice, especially on Intel CPUs.
|
|
|
|
| |
TODO: check connectivity of huts that lack doors.
|
|
|
|
| |
This seems to make it run a good deal faster.
|
| |
|
|
|
|
| |
If there are multiple, choose one randomly.
|
| |
|
|
|
|
|
|
|
| |
Sounds better, and is more consistent with boots of running.
Also, the distinction between "potion of flight" vs "boots of flying"
conveys the difference: a short-timed action vs a continuous effect.
|
|
|
|
|
|
| |
One in particular may need further purging: there's logic in greedy_explore
to skip picking up items if the player is unable to. That's never the case
anymore...
|
|
|
|
|
| |
More potential crashes to test, Lugonu does nothing interesting when you
don't use her abilities.
|
|
|
|
| |
Not that it makes any difference...
|
| |
|
|
|
|
|
|
|
| |
Unlike, say, miscasts, there is little randomness there: a kraken dies
either to a blow to its head, tentacle, or tentacle fragment, multipled
by bleeding/not bleeding, or the living or spectral one. No need to
waste much CPU repeating it over and over.
|
|
|
|
| |
Also, announce when they are finished.
|
| |
|
|
|
|
|
|
|
|
|
| |
It has been broken for more than two years, and never worked right in the
first place (ie, a bug, not a regression). It would be good to fix it, but
there's plenty of far worse bugs rotting on Mantis.
I'm disabling this one so it's possible to do automated regression tests
from a cron job.
|
|
|
|
|
|
| |
A yet another crasher has just been cherry-picked to 0.11 without anyone
noticing.
(This is just a test case, not a fix.)
|
|
|
|
| |
Let's hope no one removes artificers soon...
|
| |
|
| |
|
|
|
|
|
|
|
|
| |
It uses a temporary file for the save file. This fixes canned tests breaking
if a run has been aborted (and thus a save by that name exists), and also
allows concurrently running multiple instances of the same test.
Implies wizmode.
|
|
|
|
|
| |
The starting AK Abyss has several special properties, including never
generating any vaults as they are "scary".
|
|\ |
|
| | |
|