For the attached example I get:
decl.f90 line 3:190 no viable alternative at input ''
This is probably related to bug 3306187 but I can't modify that one so here is a new one instead.
This problem occurs with version:
commit 5a4278e497d93fa13f7de87066d7fdd87968bd88
Author: rasmussn <rasmussn@kirkelanlgov.local>
Date: Tue Jun 14 00:11:22 2011 -0600
test input
I have another example I came across that I remember gave me head aches in our own parser when there are comments following the continuation character like this:
subroutine foo()
500 FORMAT ( &!CORE..72
& '0 ICH L FLOCH1 C0FLCH C1FLCH C2FLCH',&
& ' C3FLCH TEXPEL ENVAPR DMCHPR')
600 FORMAT (1X,2I6,8D13.5) !CORE..75
700 FORMAT (' PIN3= ',D13.5,' PX3= ',D13.5,' DTSUB= ',D13.5,/, &!CORE..76
& ' ICH COEFE0 COEFE1 COEFE2 CHFLO3 ', &
& 'GAMCH THT2CH DUM1V DUM2V DEN1V', &
& ' DEN2V ',/,(I3,6D13.5,/,3X,10D13.5))
END SUBROUTINE COREFL
end subtourine
Sorry, I would have done it as an attachment but it looks like I am not allowed.
anod now - of course - this example lost the formatting ... argh
The bug was caused by an inconsistency between the count and the index in FortranStream. It was fixed by returning a 'tuple' of (index,count) rather than attempting to calculate the count from the index change.