possible bug in process.c?

possible bug in process.c?

[Michael Vezie]

In process.c, in DEFUN("start-process...

Around line 1193 or so (in 21.3, although I've seen the same code in 20.7) is the following 
snippet of code:

  /* If program file name is not absolute, search our path for it */
  if (!IS_DIRECTORY_SEP (XSTRING (program)->data[0])
      && !(XSTRING (program)->size > 1
           && IS_DEVICE_SEP (XSTRING (program)->data[1])))
    {

...  at which point it then processes a relative pathname file (looking through Vexec_path
for the executable, etc).

Maybe I'm reading this wrong (help me out if I am), but shouldn't that last line there read:
           && !IS_DEVICE_SEP (XSTRING (program)->data[1])))
              note  ^

Here's how I read it:

If the first character is not a directory separator ('/' or '\\') and the string is longer than 
1 and the second character is a device separator (':')...  you're checking here if the string denotes 
a relative path.  But if the second character is a ':', then it's something like:

c:/fsf/emacs-21.3/bin/cmdproxy.exe

which is an absolute path, and doesn't require looking through Vexec_path.

In any case, I get an error preceeded by "Searching for program" when I set my SHELL environment 
variable to c:/fsf/emacs-21.3/bin/cmdproxy.exe (it should never search for the program in the path 
with an absolute path like that), and things work fine when I set my SHELL environment variable to 
simply cmdproxy.exe

Michael

Re: possible bug in process.c?

[Eli Zaretskii]

> Date: Wed, 14 Jan 2004 13:52:44 -0500
> From: Michael Vezie <mlv@pobox.com>
> 
>   /* If program file name is not absolute, search our path for it */
>   if (!IS_DIRECTORY_SEP (XSTRING (program)->data[0])
>       && !(XSTRING (program)->size > 1
>            && IS_DEVICE_SEP (XSTRING (program)->data[1])))
>     {
> 
> ...  at which point it then processes a relative pathname file (looking through Vexec_path
> for the executable, etc).
> 
> Maybe I'm reading this wrong (help me out if I am), but shouldn't that last line there read:
>            && !IS_DEVICE_SEP (XSTRING (program)->data[1])))
>               note  ^
> 
> Here's how I read it:
> 
> If the first character is not a directory separator ('/' or '\\') and the string is longer than 
> 1 and the second character is a device separator (':')...  you're checking here if the string denotes 
> a relative path.  But if the second character is a ':', then it's something like:
> 
> c:/fsf/emacs-21.3/bin/cmdproxy.exe
> 
> which is an absolute path, and doesn't require looking through Vexec_path.

My reading of that code is:

 If the first character is not '/' or '\\' and either the file name
 is not longer than 1 character or its second character is not ':',
 then it's not an absolute file name.

> In any case, I get an error preceeded by "Searching for program" when I set my SHELL environment 
> variable to c:/fsf/emacs-21.3/bin/cmdproxy.exe (it should never search for the program in the path 
> with an absolute path like that), and things work fine when I set my SHELL environment variable to 
> simply cmdproxy.exe

Could you please see with a debugger whether the code in question is
indeed the one that's misbehaving?