Re: Compile issues with xplain2sql

On 12/21/2017 22:45, Berend de Boer wrote:
> For a long time I haven't been able to compile xplain2sql with
> gobo. The error messages do not really indicate what to do to fix the
> problem.
> 
> Here's the output:
> 
> xplain2sql/src# geant compile_ge
> Parser contains 1 shift/reduce conflict and 3 reduce/reduce conflicts.
> 
> [CATCALL] class SQL_GENERATOR_TSQL65 (ANY,95,8): type 'STRING_8' of 
> actual argument #1 does not conform to type 'UC_STRING' of formal 
> argument in feature `is_equal' in class 'UC_STRING'
[...]

> The location of the error is irrelevant it seems, it complaints about
> the ANY class. But I can't find the actual problem.

Catcall errors detected through dynamic type set analysis are not easy
to report because they depend on several factors. From the errors,
I can see that in class SQL_GENERATOR_TSQL65 (or one of its ancestors)
you are calling `equal (x, y)' where `x' might be attached to an
object of type UC_STRING and `y' may be attached to an object of type
STRING. I can see that you make many of such calls in class
SQL_GENERATOR. I suggest that you replace these calls by
`STRING_.same_string (x, y)' where `STRING_' is inherited from
KL_IMPORTED_STRING_ROUTINES.

Note that the catcall errors reported by 'gec' are just potential
catcalls. If you can compile without such errors you are guaranteed
that no catcall will occur at runtime. But even if you get such
errors, it does not mean that catcall will occur at runtime.
So, alternatively you can use the compilation option --catcall=no
when compiling with 'gec'. It will not report catcall errors at
compile time, but will raise an exception if one occurs at runtime.
(I didn't check, but this option might only be support on the
command-line, not through Xace. We might be able to configure it
in eiffel.eant though, with some variable instead of hard-coding
'catcall="error"'.)

> I have been unable to enable catcalls in ISE Eiffel, I don't think
> there is an option for that in system.xace.

The catcall detection mechanism in ISE Eiffel is different than the
one in Gobo. It's not based of dynamic type set analysis but on the
use of keywords when declaring Eiffel entities in the class text.
It makes error messages easier to understand, but it will reject
more code that with Gobo and it requires to edit nearly all classes
to add these keywords (same kind of effort at the one to convert
a non-void-safe library to void-safety).

-- 
Eric Bezault
mailto:[email protected]
http://www.gobosoft.com

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot