Re: how to use a different /bin/sh with GNU Make?

[Mark Galeck]

>All you have to do is change the list of UNIX shells to be empty:

diff --git a/job.c b/job.c
index 244def0..4b345b6 100644
--- a/job.c
+++ b/job.c
@@ -423,6 +423,7 @@ is_bourne_compatible_shell (const char *path)
 

???  Forgive me, but I made the change to empty list of unix shell (to the git 
version of Make 4.1 with Paul's patch to take care of NULL ttyname) .  Then, 
each simple recipe command, like, "touch foobar", goes through the line 
execvp (argv[0], argv);

(around line 2500 of job.c)
Here argv is the array "touch" "foobar" .  The shell /bin/sh does not know 
about this call to execvp at all...


In case of a more complicated command, the situation appears even more strange. 
 If a recipe command is 
touch foobar; touch foobar1

then we call execvp as well, this time on the array of words of
"bin/sh -c touch foobar; touch foobar1"
This is strange, because here, /bin/sh will know about "touch foobar" but will 
not know about "touch foobar1".  


I want to apologize - this thread was started because I thought, the manual 
says that for each recipe line, a new shell process is spawned.  Actually, the 
manual says that, but later in parenthesis, in the next sentence, says "in 
practice, we take shortcuts that do not affect the results).  I never read the 
second sentence...  I mean, perhaps the whole paragraph should be constructed 
better, but I should have read everything.  
I am just writing this for the record so someone else who tries to plug a 
different /bin/sh and finds this thread, is not stumped.  
Mark