[Submitted by J. David Bryan, jdbryan at acm dot org]

Thanks, but we've already had this discussion:

http://lists.gnu.org/archive/html/make-w32/2006-03/msg00057.html

Please disregard comment #2; I'll continue to apply a local patch.

[Submitted by J. David Bryan, jdbryan at acm dot org]

Would you consider adding the "move," "popd," "pushd," and "start" built-in commands as well?

Hello,  I asked for help on this issue (with GNU make on Windows) on the help-make list, and Paul Smith says, this is a bug.  I am using WinXP Pro, SP3.  Here is an SSCCE: 

makefile:
-------------------------
SHELL=cmd.exe
foobar:
@echo.>foo
------------------------

gives me:
C:\tmp>make foobar

C:\tmp>


However, this one:

makefile:
-------------------------
SHELL=cmd.exe
foobar:
@echo.
------------------------

gives:

C:\tmp>make foobar
process_begin: CreateProcess(NULL, echo., ...) failed.
make (e=2): The system cannot find the file specified.
make: * [foobar] Error 2


Here's what Paul says: 

"

I'd guess that this is due to make's fast path processing: if make can determine that the command you're invoking does not need a shell, it won't start one; instead it will simply try to invoke the command directly.

In this case, the command "echo." is not known to make as a command that requires a shell, so it tries to run it directly.  In reality this is not a real command but rather a builtin command for the Windows shell (for example command.com), so trying to invoke it directly fails.

On the other hand, when you run "echo.>foo" make sees the redirection
(">") and understands that this is not a simple command and can't be run using the fast path, so it invokes the shell to run it and it works.

(...)

It's not in the manual because it's supposed to be invisible to the user (that is, it's an implementation detail, not a user-visible feature, and so not documented in the user manual).  The idea is to improve performance by avoiding an extra (and sometimes costly) shell invocation in situations where it can be shown that the results are the same either way.

If it's visible to the user, as it is in this case, then there's a bug in the fast path processing.

" (end of quotes from Paul)


Mark