Comment #4 on issue 3439 by
[hidden email]: Compiler
error 'subscript is out of bounds: ...'
http://code.google.com/p/pharo/issues/detail?id=3439Ah Ah, I posted previous comment too fast, because I just remembered one
simple clean-up idea I had tonight that produces just the effects I would
have expected :)
Together with a protection of insertion point in SyntaxError, that should
solve the issue once ported to Pharo (despite the old compiler might be a
few versions off).
Nicolas Cellier uploaded a new version of Compiler to project The Trunk:
http://source.squeak.org/trunk/Compiler-nice.225.mcz==================== Summary ====================
Name: Compiler-nice.225
Author: nice
Time: 22 February 2012, 2:57:07.872 pm
UUID: 77bbb3ac-e939-4975-a139-f785dc94c17a
Ancestors: Compiler-nice.224
Remove the ugly hack in Parser>>expected: which was modifying the error
message insertion mark depending on the fact that we were at end of token
stream a,d/or end of source character stream.
Replace it with a proper mark in Scanner>>scanToken that correctly take
into account the two step of advance handled by the Scanner (stateful is
harmful and require high precision surgery).
While at it, set the error insertion mark after the $ character when
offended by Scanner>>xDollar.
This efficiently reduce the number of failures reported by
CompilerNotifyngTest in its Tests-nice.144 version.
=============== Diff against Compiler-nice.224 ===============
Item was changed:
----- Method: Parser>>expected: (in category 'error handling') -----
expected: aString
"Notify a problem at token 'here'."
- tokenType == #doIt ifTrue: [hereMark := hereMark + 1].
- hereType == #doIt ifTrue: [hereMark := hereMark + 1].
^ self notify: aString , ' expected' at: hereMark + requestorOffset!
Item was changed:
----- Method: Scanner>>scanToken (in category 'expression types') -----
scanToken
[(tokenType := self typeTableAt: hereChar) == #xDelimiter]
whileTrue: [self step]. "Skip delimiters fast, there almost
always is one."
mark := aheadChar == DoItCharacter
+ ifTrue: [hereChar == DoItCharacter
+ ifTrue: [source position + 1]
+ ifFalse: [source position]]
- ifTrue: [source position]
ifFalse: [source position - 1].
(tokenType at: 1) == $x "x as first letter"
ifTrue: [self perform: tokenType "means perform to compute
token & type"]
ifFalse: [token := self step asSymbol "else just unique the
first char"].
^token!
Item was changed:
----- Method: Scanner>>xDollar (in category 'multi-character scans') -----
xDollar
"Form a Character literal."
+ aheadChar == DoItCharacter
+ ifTrue:
+ [mark := mark + 1. "Let the notification lie behind
the dollar"
+ ^self offEnd: 'A Character was expected'].
- aheadChar == DoItCharacter ifTrue: [^self offEnd: 'A Character was
expected'].
self step. "pass over $"
token := self step.
tokenType := #character!
_______________________________________________
Pharo-bugtracker mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-bugtracker
Locked
