Curses feature branch

[AnonymousHero]

Nick--
I looked at this a little more. As far as I can tell, this says almost nothing about what is happening. By the time this fires, the memory has been freed, and initialized by the debug malloc package so that everything is the value 0xfeeefeee repeated. (Other allocators may use 0x0badbeef instead.)
When I get a chance I'll use fsanitize.

Sanitize is really the only option here. Usually it'll tell you the exact line in the source that did "the bad thing".

Last time I tried, the Angband makefiles required quite a bit of hackery to get -fsanitize=XXX working... because, well, they're a little bit broken by modern standards. (FWIW, my offer to convert Vanilla Angband to use CMake still stands... ultimately it *is* really just easier than the hell of autoconf/automake.)

[Nick]

Nick--
I looked at this a little more. As far as I can tell, this says almost nothing about what is happening.

Not without context, but both those stack traces look familiar

The new regime - objects as linked lists on each grid and a whole player known copy of the current level - is nice as a model of what's going on, but it is taking time to distinguish all the cases of what memory should be freed when.

[Pete Mack]

fsanitize is very cool. But purify was even cooler: you didn't need a compiler, just a debug build. It instrumented object code directly, so there was no need for messing with makefiles, or picking a new compiler, etc.

[AnonymousHero]

fsanitize is very cool. But purify was even cooler: you didn't need a compiler, just a debug build. It instrumented object code directly, so there was no need for messing with makefiles, or picking a new compiler, etc.

I can't claim to have had any experience with purify(TM, I'm assuming), but I take it from your use of the past tense that it's not some magic(?!?) you still have access to? Sanitize is ridiculously good, btw, and any lesser superlative is insufficient .

[Pete Mack]

Purify licenses are ridiculously expensive since IBM bought Rational, so no: don't have access. But yes, Sanitize is just as good--just slightly harder to use, since you may need a new build process, rather than just executing it as part of the command line.

[Nick]

Speak now

My current plan is to pull this branch into the master branch soonish. If you have an objection to the way curses are done here, please voice it.

[Ingwe Ingweron]

My current plan is to pull this branch into the master branch soonish. If you have an objection to the way curses are done here, please voice it.

Several bugs still not addressed, including an occasional crash bug, and others I've reported above. Have provided savefiles above for some, but still need to get save files for others.

1. Sometimes pass-through walls monsters are dropping their treasures inside walls. Savefile was provided. Don't remember it being addressed, but I haven't been focusing on whether it's still an issue.

2. Object indicator ("*") of previously found and ignored items reappearing when within a new detection radius. Savefile was provided. This is definitely still happening.

3. Still an intermittent crash bug that I haven't been able to run down, though the Plain Gold Ring one (and Morgoth's crown) is fixed.

4. Still have the bugs I reported but haven't gotten a savefile for yet. Haven't had as much time to play, but I know the gobbly-gook message on remove curse one still occurs intermittently. Not sure about the others.

[Nick]

Several bugs still not addressed, including an occasional crash bug, and others I've reported above.

Yes, I haven't got to those (or anything else) yet.

I was more after whether the general approach is OK, so I know if this branch (with bugs fixed) is worth persisting with.

[Huqhox]

I'll vote yes.

Curses are annoying for the player but add dimension to the gameplay and I'm glad to see them back and with some nice ideas and features.

It's not perfect yet and there's a way to go but they will make 4.1 a far more interesting gaming experience.

[Nick]

This branch has now been incorporated into master - builds are appearing an the nightlies page as I type. This does not mean the bugs have been fixed

[PowerWyrm]

This branch has now been incorporated into master - builds are appearing an the nightlies page as I type. This does not mean the bugs have been fixed

Great, I'll have something to look at before my holidays

[PowerWyrm]

Well, there's something off about the curses branch when it comes to artifacts.

@ just found the Hard Leather Armor of Himring (+0,+0) [20,+15], which has poison resistance, but in this case is also cursed with poison. (Poison power 10, when @ casts remove curse.) No strange characters in the message line this time, though. Also, I don't recall body armor having pluses to hit and damage identified (even if they are zero). I'm not talking about an encumbrance penalty, but the (+0,+0) on Himring.

I just rapidly looked at the changes, and there's nothing in the code that prevents artifacts from getting a curse: you still get the 5% chance of a curse being added like normal objects.

[Nick]

I just rapidly looked at the changes, and there's nothing in the code that prevents artifacts from getting a curse: you still get the 5% chance of a curse being added like normal objects.

I guess that had better come out

[PowerWyrm]

Also I don't really understand the poison and paralysis curses, since they can be countered by resist poison and free action. Replace poison by sickness (same as poison but not resistable) and paralysis by fainting (same as the salt water effect) and they would be fine.

[Pete Mack]

Both of those are too lethal. They make the item unusable, especially fainting. Poison can be worked around with a spell.