On Wed, Nov 21, 2018 at 11:44:06PM +0100, Antonio Ospite wrote:
> On Wed, 21 Nov 2018 22:27:16 +0100
> Luc Van Oostenryck <luc.vanoostenryck@xxxxxxxxx> wrote:
>
> > Currently cgcc only process file given as stdin or if their
> > name end with '.c'. Other files are explicitly ignored.
> >
> > This generally correspond to what is wanted but GCC also accept
> > files with any extension if their type is given via the option
> > '-x <language>'. Some projects use this mechanism, for example
> > to use the C pre-processor on some files containing no code.
> > This fails when cgcc is used as wrapper around sparse + GCC.
> >
> > Fix this by teaching sparse about the '-x c' option.
> >
> > Reported-by: Antonio Ospite <ao2@xxxxxx>
> > Signed-off-by: Luc Van Oostenryck <luc.vanoostenryck@xxxxxxxxx>
> > ---
> > cgcc | 8 ++++++++
> > 1 file changed, 8 insertions(+)
> >
> > diff --git a/cgcc b/cgcc
> > index 7611dc9ff..188197369 100755
> > --- a/cgcc
> > +++ b/cgcc
> > @@ -25,6 +25,14 @@ while (@ARGV) {
> > # Ditto for stdin.
> > $do_check = 1 if $_ eq '-';
> >
> > + # accept any file type input language is specified with '-x c'
> > + if ($_ eq '-x') {
> > + my $lang = shift(@ARGV);
> > + die ("$0: missing argument for -x") if !$lang;
> > + $do_check = 1 if $lang eq 'c';
> > + next;
> > + }
> > +
>
> Hi Luc,
>
> this improves things a bit, but there is another issue with -E.
>
> If "-o file" is passed and only $do_check is enabled sparse will
> ignore the -o option and the preprocessed output will go to the stdout
> and not in the specified file, as would be expected by a normal gcc
> invocation.
Yes, I know. Sparse always ignored the '-o option' but I added a patch
for it yesterday.
> So the question is: does it make sense to run sparse on files which
> are only preprocessed?
It really does make sense. The biggest reason is because you want
to have coherent results between the output of '-E' and sparse's
pre-processing at checking time (for example, if the code contains
#ifdef __CHECKER__).
I can also imagine cases where it make less sense, for example
because '-E' need to produce an output and sparse (when simply
used as a checker) doesn't, like in your case here.
> Or can -E be treated in the same way as -M is
> handled?
The -M is handled so because sparse doesn't have the support for it
so its delegated to $(CC).
> I really don't know as I don't work a lot with sparse, I am just a
> casual user.
>
> I was thinking about something like:
>
> diff --git a/cgcc b/cgcc
> index 7611dc9..3a00819 100755
> --- a/cgcc
> +++ b/cgcc
> @@ -9,6 +9,7 @@ my $m32 = 0;
> my $m64 = 0;
> my $has_specs = 0;
> my $gendeps = 0;
> +my $preprocess = 0;
> my $do_check = 0;
> my $do_compile = 1;
> my $gcc_base_dir;
> @@ -28,6 +29,7 @@ while (@ARGV) {
> $m32 = 1 if /^-m32$/;
> $m64 = 1 if /^-m64$/;
> $gendeps = 1 if /^-M$/;
> + $preprocess = 1 if /^-E$/;
>
> if (/^-target=(.*)$/) {
> $check .= &add_specs ($1);
> @@ -52,9 +54,6 @@ while (@ARGV) {
> next;
> }
>
> - # If someone adds "-E", don't pre-process twice.
> - $do_compile = 0 if $_ eq '-E';
> -
> $verbose = 1 if $_ eq '-v';
>
> my $this_arg = ' ' . "e_arg ($_);
> @@ -62,7 +61,7 @@ while (@ARGV) {
> $check .= $this_arg;
> }
>
> -if ($gendeps) {
> +if ($gendeps || $preprocess) {
> $do_compile = 1;
> $do_check = 0;
> }
>
>
> If this is acceptable I can send a proper patch.
If fact, I like it because it uses a more 'positive logic'
but it would change the current behaviour. So, if the patch
with '-o' works for you, I would prefer to leave it as is
for now.
Thanks,
-- Luc
