This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: Testsuite bugs (corrected)


On 19 Mar 1999 10:36:02 +0100, Andreas Schwab wrote:
>Zack Weinberg <zack@rabi.columbia.edu> writes:
>
>|> On 18 Mar 1999 02:59:11 +0100, Andreas Schwab wrote:
>|> >Zack Weinberg <zack@rabi.columbia.edu> writes:
>|> >
>|> >|> On 17 Mar 1999 10:40:50 +0100, Andreas Schwab wrote:
>|> >|> >This fixes the tests gcc.dg/cpp-if1.c and gcc.dg/cpp-if3.c for real so
>|> >|> >that they don't spuriously fail.  The last patch i sent was not uptoda
>te,
>|> >|> >unfortunately.
>|> >|> 
>|> >|> ...
>|> >|> 
>|> >|> >-#if U_MAX == ULL_MAX || LL_MIN == 0 || LL_MAX == -1
>|> >|> >+#if U_MAX == ULL_MAX || LL_MIN == 0 || LL_MAX == -1 /* { dg-error "to
>o ma
>|> >ny `
>|> >|> >l'" "too many suffixes" } */
>|> >|> 
>|> >|> This line should not be generating an error message.  C9x allows two
>|> >|> `l' suffixes on an integer constant, and when I tested it both cccp
>|> >|> and cpplib got this right.
>|> >
>|> >The default for -pedantic is still C89.  What did you test??  See cccp.c:
>|> >
>|> >	    if (!pedantic < spec_long)
>|> >	      yyerror ("too many `l's in integer constant");
>|> 
>|> cccp -pedantic with no -lang switches.  When I tested it (it was
>|> awhile ago) that meant C9x extensions are accepted.
>
>Nowhere in the cccp source is a test that combines pedantic with c9x.  All
>pendantic tests are against the _current_ standard, independent of the
>-std setting.  IMHO this matches exactly the docs, thus ccplib is broken.

I interpret the docs to say -pedantic means be pedantic about
conformance to the standard selected with -lang and/or -std switches.
We can argue about that later.  This test is, among other things,
testing cpp's ability to do 64 bit arithmetic.  How would you make
cccp accept 64 bit numbers while still generating all the other
warnings and errors that the test checks for?  If this is not
possible, we need to split the testcase.

zw


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]