Search results

I have not given adequate thought to that PR in over a year. I saw Rinn's fix, and I think it's a very plausible explanation. I would be happier if we had better tooling to ensure that all reads / writes were properly synchronized.

Thanks! I suppose there's still a secondary bug where we flip back and forth between two values due to KoL reporting an incorrect level in api.php. We're also still not detecting level from charpane.php for the titles that embed %L and don't actually mention the word "level", although I am not...

Of course, the character I'm testing with doesn't have charpane.php with the default UI detect the level (custom title or not), since we apparently only handle the cases of compact charpane and regular charpane with a custom title that doesn't contain %L (where the level renders after the title...

Ah, yes, r28372 would have done it -- logged into a char with max stats, and api.php shows level 255. Meanwhile r28391 is fighting with it every time charpane.php is loaded, which detects the level as 256, but the next load of api.php resets the level back to 255 again.

No, the max level has been 256 for some time. I suspect that the new logic re: level parsing in r28391 broke this. edit: or possibly r28372.

Ha. Yes, there are lots of advantages, like for instance durability (and the rest of the ACID properties). The main "disadvantage" is that it is less human-editable. Perhaps the biggest thing stopping us from adopting this is migration cost. I had previously attempted to start a discussion on...

Both can be true! Note that the code returned early if you were running a daily build (`KoLConstants.VERSION_NAME.startsWith( "KoLmafia r" )`). https://github.com/kolmafia/kolmafia/blob/a0b5d8ad4d0657a4f8e151152b50fbdf31060bee/src/net/sourceforge/kolmafia/KoLmafia.java#L2152 Incidentally, the...

I believe it dates back to the days of point releases (which we differentiated from the daily builds via "version" vs "revision"). If there was a newer point release than what you were running, Mafia would create a popup recommending that you update. As I'm sure you recall, some segment of our...

Mafia currently thinks your mainstat is muscle. My guess is that KoL is treating your mainstat as the max of the three, similar to Plumber or Grey Goo, but we need different special handling for determining your level in this path.

The error that I ran into (and which I assume fewyn ran into) was in a script that essentially invoked sell_price($item[cyburger].seller, $item[cyburger]). sells_item unfortunately does not help here, although I think you're spot on with your observation of coinmasters with multiple costs.

I'm inclined to agree, although maybe for slightly different reasons. One, I think JSON is overkill for what we need. Two, I would prefer to keep our external dependencies to a minimum. log4j in particular burned the entire industry quite badly a few years back, so I'm starting from a point of...

Thanks @Ryo_Sangnoir for r28355 (adapting my patch from #4), which should at least stop the NPEs until we have time to investigate further what the desired behavior should be.

My somewhat hacky workaround without fully understanding why CoinmasterData.token is null is to add null checks in isHP() and isMP(). (And for completeness, the other similar methods that dereference a string without checking if it's null first). diff --git...

Ah, I see fewyn and soolar both reported variants on this in https://kolmafia.us/threads/shop-php-support.30296/#post-176321

Looks like this was introduced in r28347 -- doesn't reproduce in r28346.

As far as I can tell, this started happening in the past ~day. Minimal reproducer: ash sell_price($coinmaster[the dedigitizer], $item[cyburger]) Stack trace that I got from my script that was crashing: class java.lang.NullPointerException: Cannot invoke "String.equals(Object)" because...

I can see the entire private repository because I am part of the KoLmafia organization. I cannot push to the repository, or fork it. It is not publicly viewable.

Might be a matter of deleting the toolchain { ... } stanza which makes more sense if we're raising the compiler floor but isn't necessary if we're sticking to an old version.

Exactly -- that's an authentication issue. Often, for security reasons it's preferable to return Not Found instead of, say, Unauthorized so you don't leak side channel information (in this case: whether a specifically-named repository exists)...

Apparently this is a long-standing issue with Github Desktop -- see https://github.com/desktop/desktop/issues/3870 Just run the following commands: $ git update-index --chmod=+x gradlew $ git commit -m "Update file permissions for gradlew"