[PATCH] mixing of labels and code in C2X

[Joseph Myers]

On Sun, 13 Sep 2020, Uecker, Martin wrote:

> Hi Joseph,
> 
> here is the (unfinished) patch to support
> mixing of labels in C2X.

I think there should be explicit tests for old standard versions 
(c11-labels-1.c etc.) that these usages are errors with -pedantic-errors 
with the old -std option, are warnings with -pedantic, and are quietly 
accepted with neither.  In addition to using -pedantic-errors with the new 
standard version test to confirm it's not diagnosed, and a test with the 
new version and -Wc11-c2x-compat.

By way of example of further places that I think need changing in the 
patch: at present, c_parser_label gives an error (that you correctly 
change to a pedwarn_c11) if the label is followed by a declaration - and 
then parses the declaration itself rather than leaving it to be parsed in 
the caller.  So c_parser_compound_statement_nostart would parse a label 
followed by a declaration, and at that point last_label would be set to 
true, meaning that a second declaration would be rejected, when in C2x it 
should be accepted.  You can see this even without the patch with a test 
such as:

void
f (void)
{
 label : int a; int b; 
}

I think that instead c_parser_label should never try to parse a 
declaration following the label; that should be left for the caller to 
handle.  To avoid c_parser_label needing to return information about 
standard attributes on a following declaration, maybe it shouldn't parse 
standard attributes either (note that means that c_parser_all_labels would 
need to handle parsing and warning about and discarding standard 
attributes after each label instead - such attributes might be ones on a 
statement, or ones on the next label in a sequence of labels).

Then of course the checks of last_label in 
c_parser_compound_statement_nostart would need to become checks for 
whether to pedwarn_c11 about the use of a label in a given context, once 
the code knows whether there is a declaration, rather than preventing 
parsing a declaration there at all.  So probably c_parser_label would no 
longer have the pedwarn_c11 either; that would all be left to its callers.

-- 
Joseph S. Myers
joseph@codesourcery.com