[Bug other/96287] New: Empty string argument to gcc should be ignored

[Bug other/96287] New: Empty string argument to gcc should be ignored

[andreas.urban at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96287

            Bug ID: 96287
           Summary: Empty string argument to gcc should be ignored
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: other
          Assignee: unassigned at gcc dot gnu.org
          Reporter: andreas.urban at hotmail dot com
  Target Milestone: ---

Passing an empty string to GCC makes it look for a file named as such:

$ gcc --version
gcc (Debian 6.3.0-18+deb9u1) 6.3.0 20170516
$ gcc
gcc: fatal error: no input files
compilation terminated.
$ gcc ""
gcc: error: : No such file or directory
gcc: fatal error: no input files
compilation terminated.

This is inconvenient for shells invoking other programs with a list of
arguments, such as Perl, Ruby, and Tcl, as you would have to remove the empty
strings from the argument list before executing command.

My suggestion is to have gcc ignore empty string arguments instead of treating
it as a filename. This will make integration with said languages and others
simpler at almost no cost.

Since empty string is not a supported filename for any mainstream filesystem,
and a terrible choice of name where supported, I see no apparent risk of
compatibility problems.

"Define errors out of existence" - John Ousterhout

[Bug other/96287] Empty string argument to gcc should be ignored

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96287

--- Comment #1 from Jonathan Wakely <redi at gcc dot gnu.org> ---
Ignoring it could lead to equally undesirable behaviour though.

for file in *.cc ; do gcc "$fil" ; done

Don't those languages support something like the Bourne shell's "$@" which does
the right thing?

[Bug other/96287] Empty string argument to gcc should be ignored

[schwab@linux-m68k.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96287

--- Comment #2 from Andreas Schwab <schwab@linux-m68k.org> ---
If you pass a non-file-name where a file name is expected you are doing
something wrong, and you need to fix *that*.  Hiding errors is doing a
disservice.

[Bug other/96287] Empty string argument to gcc should be ignored

[andreas.urban at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96287

--- Comment #3 from Andreas Urban <andreas.urban at hotmail dot com> ---
(In reply to Jonathan Wakely from comment #1)
> Ignoring it could lead to equally undesirable behaviour though.
> 
> for file in *.cc ; do gcc "$fil" ; done
> 
> Don't those languages support something like the Bourne shell's "$@" which
> does the right thing?

Tcl has {*} expanding a list to be treated as arguments. Empty elements
disappear. Thinking about it some more, yes that behaves as intended and is
relatively easy to read and write:

exec gcc {*}$cppflags {*}$cflags -o $@ $<

So we can cross out Tcl from the list of languages that benefit from such
change.

[Bug other/96287] Empty string argument to gcc should be ignored

[andreas.urban at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96287

--- Comment #4 from Andreas Urban <andreas.urban at hotmail dot com> ---
Looking closer at how Perl exec works, along with join on empty strings and
variable, there would seem to be no problem:

exec 'gcc', join(' ', @cppflags, @cflags, '-o', '$@', '$<')

There may be one or two things off in above line but from my experiments on
join and exec it behaves as one would want for empty arrays.

I'm now convinced this is not a big issue. Closing issue.

[Bug other/96287] Empty string argument to gcc should be ignored

[andreas.urban at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96287

Andreas Urban <andreas.urban at hotmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |INVALID
             Status|UNCONFIRMED                 |RESOLVED

--- Comment #5 from Andreas Urban <andreas.urban at hotmail dot com> ---
Withdrawing proposal. Alternative solutions found satisfactory.