(Also submitted as BugID 7606)
(Platform: QNX 6.2.0)
* When a "test" command is a shell builtin, and
* no /bin/test as a file is provided in the system,
all:
test
will fail with "make: test: Command not found" .
You could easily reproduce this by setting PATH to bogus path, as
$ PATH=/no
%% <[EMAIL PROTECTED]> writes:
k> POSIX says the line should be processed like system("test")
k> which will be execvp("/bin/sh","-c","test") which will succeed,
k> but GNU make tries to optimize avoiding /bin/sh,
k> which tries execvp("test") and fails.
k> GNU automake uses this bare "
[EMAIL PROTECTED] sed in <[EMAIL PROTECTED]>:
>> k> GNU automake uses this bare "test" in several testsuite.
>>
>> I don't have a problem making this change but I'm curious as to what
>> kind of command line in a makefile rule would use test(1) but not use
>> any other special character that wo
[EMAIL PROTECTED] sed in <[EMAIL PROTECTED]>:
>> any other special character that would cause make to kick the command to
>> the slow path?
BTW Solaris /bin/test is a shellscript which is merly
`basename $0` "$@"
so on platforms like this, the slow path (sh -c test) should be faster.
Well Linux a
%% <[EMAIL PROTECTED]> writes:
>>> any other special character that would cause make to kick the command to
>>> the slow path?
k> BTW Solaris /bin/test is a shellscript which is merly `basename $0`
k> "$@" so on platforms like this, the slow path (sh -c test) should
k> be faster.
Why f
[EMAIL PROTECTED] sed in <[EMAIL PROTECTED]>:
>> %% <[EMAIL PROTECTED]> writes:
>>
>> >>> any other special character that would cause make to kick the command to
>> >>> the slow path?
>>
>> k> BTW Solaris /bin/test is a shellscript which is merly `basename $0`
>> k> "$@" so on platforms
