[gfortran,patch] Internationalisation of the Fortran front-end

Tobias Schlüter tobias.schlueter@physik.uni-muenchen.de
Fri Sep 2 09:35:00 GMT 2005


Steve Kargl wrote:
> On Fri, Sep 02, 2005 at 03:45:03AM +0200, Tobias Schl?ter wrote:
> 
>>Tobias Schl?ter wrote:
>>
>>>What exactly are the advantages of using the common machinery if the most
>>>complicated part of the code (preparing the lines for caret printing and
>>>printing them) will still have to be done in the frontend (and probably
>>>reimplemented, as gfortran's code will probably not drop into place)?
>>
>>Well, coming to think of it, there are other complications in gfortran's error
>>printing code: error messages are buffered, so that if one matcher determines
>>that the statement is not a correct instance of the statement it's trying to
>>match, a later matcher still can accept the statement without an error message
>>being printed.  I have no idea how well this mechanism would fit in with the
>>common machinery.
>>
> 
> 
> OTOH, gfortran's error machinery can get awfully confused
> where a single error can lead to cascade of errors that
> eventually wedge gfortran.
> See PR 23538.  

I haven't analyzed this specific cascade of errors, but the parser is
deficient, in that it gets very easily confused if a statement beginning or
ending a block contains an error.  Every following subroutine or function
statement is then rejected, leading to a whole cascade of errors.  This is a
problem with the parser, not with error printing.

- Tobi
> 



More information about the Gcc-patches mailing list