Factor/To do/Bugs

Do not edit this list. If you find a bug in Factor, report it to Slava Pestov.

• there is a bug in PEGs that needs investigating. `git diff master..smalltalk extra/smalltalk/parser`
• running openssl tests twice fails on Linux/32
• custom equal? method on singletons breaks system hashtables; use id hashes where appropriate
• problem with forgetting class+generic with the same word
• erg: "i probably reported this before, but i was moving around extra/cpu/ and found a refresh issue. extra/cpu is loaded in memory and is my attempt to fix it for new accessors. i decide to try the old one, and do mv extra/cpu extra/cpu-new mv extra/cpu-old extra/cpu"
• related-words should say that the word-links definition-changed
• [ "/Users/erg/omg" f <monitor> next-change 2drop ] with-monitors doesn't work on mac, but [ "/Users/erg/" f <monitor> next-change 2drop ] with-monitors does
• win64 closing a window in ui segfaults

Haven't seen these for a while:

• doublec's windows SEH issue
• kernel reload fails on goo
• bsd epic fail -- uncomment those reload tests eventually
• moved a tuple between vocabs, some layouts didn't get updated?
• reloading sequences, assocs hangs sometimes?
• (flatten-class) doesn't have a method for word class error when reloading code after removing a class
• add unit tests for quotation compression. code heap references to quots might not be getting updated
• windows seh issues -- < mnestic> erg: it's definitely BTTray.exe from the widcomm driver set
• haskell: odd test failures
• terrorist: ui segfault, test failure
• goo inference of 'open' epic fail
• cutler has random 'Memory protection fault at address ffffffff'
• Generic word (classes-intersect?) does not define a method for the word class. when doing heavy vocab-renaming/forgetting/reloading. Maybe only when there are parse errors; also related is that forgotten words (dispatch engines) are showing up in changed definitions.

Fixed:

• call( and execute( inline caches need to be cleared when stuff is redefined
• this problem is back, we break an essential word, inference only finds out at optimize time. this is because alien-invoke doesn't appear in the usage of alien>string
• Bug Doug found:

: foo2 ( -- array )
   [
       [ 1 2 ] output>array
       [ 3 4 ] output>array
   ] output>array ;

• crash with -generations=2
• invoking restarts reloads a file multiple times again
• invoking restarts in stage1 fails with odd error -- put a typo in vectors.factor to repro
• quotation compression: non-optimized words may continue to refer to old quotations
• amended-use? not always set in the right scope
• alarms need to skip frames
• USE: furnace.actions loads furnace.actions twice
• top-level forms don't assert-depth anymore?
• compile errors: "In word: f" oops
• dlists don't implement clone correctly
• utf16 alien>string
• foldable and flushable doesn't work on methods
• deploy tool has no way to enable -include=unicode
• unit tests should not use hardcoded ports
• ./factor in dmg breaks
• sometimes, host classes end up in target image's class<=-cache
• help-lint doesn't notice bad $vocab-links
• 'pitfalls' cookbook article is out of date
• deploy tool clobbers values
• read-only, initial docs refer to old style "slot" syntax
• :: docs refer to closure conversion, but all forms are converted now at parse time
• save-bitmap doesn't work
• got this error again:

P" resource:basis/compiler/cfg/instructions/instructions.factor"
101: TUPLE: ##float>integer < ##unary ;
                           ^
Re-definition of { "resource:basis/compiler/cfg/instructions/instructions.factor" 101 }