Re: Anyone have a clue ...

Re: Anyone have a clue ...

[Dominique d'Humières]

Steve,

The block has been introduced by Jerry at r129685 (pr31306) along with the test gfortran.dg/argument_checking_7.f90. This test has been changed at r139425 by Daniel Kraft (pr32095 and pr34228).

'm == MATCH_YES && ‘ has been removed in a more recent revision.

Hope it helps.

Good luck,

Dominique

Re: Anyone have a clue ...

[Jerry DeLisle]

On 10/22/2015 10:29 AM, Steve Kargl wrote:
> On Wed, Oct 21, 2015 at 05:28:49PM -0700, Steve Kargl wrote:
>> what the code in lines 718-733 in decl.c are for?  If
>> I surround the block of code with #if 0 ... #endif, I
>> do not trigger a failure in the testsuite.  So, it appears
>> that the syntax error that is being tested is
>> not represented by a test case.
>>
> 
> Little svn spelunking showed that the lines in question
> and the chunk of code at the 'syntax:' label were introduced
> by r129685 on 2007-10-27.  At this time, two testcases were
> added as r129686, 2007-10-27, argument_checking_[7,8].f90.
> 
> It seems that checking the charlen was improved twice
> r139116, 2008-08-14 and r139425, 2008-08-22 with adjustments
> to argument_checking_7.f90 to account for new error messages.
> 
> So, the lines in question have been dead code since Aug 2008.
> I will be reaping those lines shortly.
> 
 Just now reading my mail, that was a long time ago. ;)

Jerry

Re: Anyone have a clue ...

[Steve Kargl]

On Wed, Oct 21, 2015 at 05:28:49PM -0700, Steve Kargl wrote:
> what the code in lines 718-733 in decl.c are for?  If
> I surround the block of code with #if 0 ... #endif, I
> do not trigger a failure in the testsuite.  So, it appears
> that the syntax error that is being tested is
> not represented by a test case.
> 

Little svn spelunking showed that the lines in question
and the chunk of code at the 'syntax:' label were introduced
by r129685 on 2007-10-27.  At this time, two testcases were
added as r129686, 2007-10-27, argument_checking_[7,8].f90.

It seems that checking the charlen was improved twice
r139116, 2008-08-14 and r139425, 2008-08-22 with adjustments
to argument_checking_7.f90 to account for new error messages.

So, the lines in question have been dead code since Aug 2008.
I will be reaping those lines shortly.

-- 
Steve

Anyone have a clue ...

[Steve Kargl]

what the code in lines 718-733 in decl.c are for?  If
I surround the block of code with #if 0 ... #endif, I
do not trigger a failure in the testsuite.  So, it appears
that the syntax error that is being tested is
not represented by a test case.

In looking at decl.c(char_len_param_value), this function
seems to be intended for enforcement of

F08: C417 (R422) A type-param-value in a char-length shall be a
     colon, asterisk, or specification-expr .

The colon and asterick are handled, but the rest of the 
char_len_param_value seems to be lacking in the requirements
for a specification-expr.

Now, if one looks in expr.c, one finds gfc_specification_expr().

-- 
Steve