From: Stephan Beyer <hidden> Date: 2016-06-15 22:45:02
Hence, the test passes also when you run "make" in t/
or when you invoke t4116-apply-reverse.sh directly,
without $TAR being set.
Signed-off-by: Stephan Beyer <redacted>
---
t/Makefile | 2 +-
t/t4116-apply-reverse.sh | 1 +
2 files changed, 2 insertions(+), 1 deletions(-)
From: Stephan Beyer <hidden> Date: 2016-06-15 22:45:02
Hi,
Miklos Vajna wrote:
On Fri, Jul 25, 2008 at 06:37:40PM +0200, Stephan Beyer [off-list ref] wrote:
quoted
Hence, the test passes also when you run "make" in t/
or when you invoke t4116-apply-reverse.sh directly,
without $TAR being set.
Thanks, I just hit this issue today. ;-)
Puh, I'm glad that you're not writing
"I've hit this issue yesterday and already sent a patch to this list" :)
Regards,
Stephan
--
Stephan Beyer [off-list ref], PGP 0x6EDDD207FCC5040F
From: Junio C Hamano <hidden> Date: 2016-06-15 22:45:02
Actually, 455a7f3 (More portability., 2005-09-30) introduced $TAR and it
is also used in t5000. cb34882 (fix t5000-tar-tree.sh when $TAR isn't
set, 2005-11-08) did the same fix you are adding, but I think both of
these fixes are in a wrong place.
I think we should do this instead. That's how SHELL_PATH is passed around
from build to all the test scripts already.
---
Makefile | 1 +
t/t5000-tar-tree.sh | 1 -
2 files changed, 1 insertions(+), 1 deletions(-)
But then TAR has to be set in test-lib.sh also, to be able to
invoke t5000 and t4116 directly, hasn't it?
Regards.
--
Stephan Beyer [off-list ref], PGP 0x6EDDD207FCC5040F