Ironruby core - Feb 2008 - MergeLocation bug

A Parser/scanner bug has been introduced in lastest revision.



We used to be able to successfully parse rubygems-1.0.1\setup.rb but it now
raises an assertion error in SourceSpan::ValidateLocations.



The problem lies in the Parser::MergeLocations method which I believe needs to
be able to deal with the case where the end span lies before the start span.
This occurs when there is a production with an empty RHS. Such a scenario should
produce a span of width 0.



I also note that the arguments to MergeLocation on line 350 of GPPG.cs have been
reversed - I''m not sure that is correct.



Cheers, Wayne.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://rubyforge.org/pipermail/ironruby-core/attachments/20080219/187e733f/attachment.html

Will look at it.

Thanks!
Tomas

From: ironruby-core-bounces at rubyforge.org [mailto:ironruby-core-bounces at
rubyforge.org] On Behalf Of Wayne Kelly
Sent: Monday, February 18, 2008 5:12 PM
To: ironruby-core at rubyforge.org
Subject: [Ironruby-core] MergeLocation bug


A Parser/scanner bug has been introduced in lastest revision.



We used to be able to successfully parse rubygems-1.0.1\setup.rb but it now
raises an assertion error in SourceSpan::ValidateLocations.



The problem lies in the Parser::MergeLocations method which I believe needs to
be able to deal with the case where the end span lies before the start span.
This occurs when there is a production with an empty RHS. Such a scenario should
produce a span of width 0.



I also note that the arguments to MergeLocation on line 350 of GPPG.cs have been
reversed - I''m not sure that is correct.



Cheers, Wayne.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://rubyforge.org/pipermail/ironruby-core/attachments/20080218/a5f5eb26/attachment.html