Re: [pph] New PPH command line and tests

[Diego Novillo <dnovillo at google dot com>]

On 11-02-16 07:03 PM, Lawrence Crowl wrote:

> The pre-tokenized header (PTH) tests that are currently in
> the gcc/testsuite/pph directory have been rewritten to the new
> command-line-driven scheme.  In the process, I've renamed tests to
> the following conventions in the prefix of the name.
>
>      [acdpxy][12]*
>      a - auxillary headers
>      c - positive tests for c-level headers and sources
>      d - negative tests for c-level headers and sources
>      p - positive tests for what would be c-level code, but which
>          due to C++ standard namespace games are not quite C level
>          tests.
>      x - C++-level positive tests.
>      y - C++-level negative tests.
>      1 - first priority feature tests
>      2 - second priority feature tests

Could you add a README file in g++.dg/pph documenting this convention? 
You rely on tcl expansion ordering to get at these files in the proper 
order, right?


> Index: gcc/strstrmap.c
> ===================================================================
> *** gcc/strstrmap.c	(revision 0)
> --- gcc/strstrmap.c	(revision 0)
> ***************
> *** 0 ****
> --- 1,181 ----
> + /* A string to string mapping.
> +
> +    Copyright (C) 2011 Free Software Foundation, Inc.
> +    Contributed by Lawrence Crowl <crowl@google.com> and
> +    Diego Novillo <dnovillo@google.com>.

I had nothing to do with this :)

> + /* A string to string mapping.
> +
> +    Copyright (C) 2011 Free Software Foundation, Inc.
> +    Contributed by Lawrence Crowl <crowl@google.com> and
> +    Diego Novillo <dnovillo@google.com>.

Likewise.


> +
> +   length = strlen (basename);
> +   strcpy (hdrbuf, basename);
> +   strcpy (hdrbuf+length, ".h");

Space around '+'.

> +   strcpy (pphbuf, basename);
> +   strcpy (pphbuf+length, ".pph");

Likewise.

> +   old_file = strstrmap_insert (include_pph_mapping, hdrbuf, pphbuf);
> +   if (old_file != NULL)
> +     {
> +       /* FIXME pph: free (CONST_CAST (void*, old_file)); */
> +       fatal_error ("PPH include %s was remapped: ", pphbuf);

What's this FIXME?  You're about to fatal_error, freeing memory is not 
really important.

> +       include = line;
> +       pph_file = tab+1;

space around '+'.

> +       old_file = strstrmap_insert (include_pph_mapping, include, pph_file);
> +       if (old_file != NULL)
> +         {
> +           /* FIXME pph: free (CONST_CAST (void*, old_file)); */
> +           fatal_error ("PPH include %s was remapped: ", include);

Same as before.


> --- 1154,1160 ----
>   # Language-specific object files shared by all C-family front ends.
>   C_COMMON_OBJS = c-family/c-common.o c-family/c-cppbuiltin.o c-family/c-dump.o \
>     c-family/c-format.o c-family/c-gimplify.o c-family/c-lex.o \
> !   c-family/c-omp.o strstrmap.o c-family/c-opts.o c-family/c-pch.o \
>     c-family/c-ppoutput.o c-family/c-pragma.o c-family/c-pretty-print.o \
>     c-family/c-semantics.o c-family/c-ada-spec.o

strstrmap.o should be in OBJS-common, actually.

Looks OK otherwise.

I'm currently finishing the move to use the LTO streamer to write out 
trees from the front end.  To test it out, I will hook it to -fpth to 
write the few trees produced during pre-tokenization.  So, I've added a 
pth testsuite directory and a dg-pth.exp.

Once I have that working, you can start using it to emit/read trees from 
pph.  We can decide later what to do with the pre-tokenization code.


Diego.