You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@apr.apache.org by bo...@apache.org on 2007/09/10 03:17:49 UTC
svn commit: r574113 - /apr/apr/trunk/test/Makefile.in
Author: bojan
Date: Sun Sep 9 18:17:48 2007
New Revision: 574113
URL: http://svn.apache.org/viewvc?rev=574113&view=rev
Log:
Any test returning non-zero should stop make
Modified:
apr/apr/trunk/test/Makefile.in
Modified: apr/apr/trunk/test/Makefile.in
URL: http://svn.apache.org/viewvc/apr/apr/trunk/test/Makefile.in?rev=574113&r1=574112&r2=574113&view=diff
==============================================================================
--- apr/apr/trunk/test/Makefile.in (original)
+++ apr/apr/trunk/test/Makefile.in Sun Sep 9 18:17:48 2007
@@ -47,11 +47,13 @@
check: $(STDTEST_PORTABLE) $(STDTEST_NONPORTABLE)
for prog in $(STDTEST_PORTABLE) $(STDTEST_NONPORTABLE); do \
./$$prog; \
- if test $$? = 255; then \
+ status=$$?; \
+ if test $$status != 0; then \
echo "$$prog failed"; \
break; \
fi; \
- done
+ done; \
+ exit $$status
occhild@EXEEXT@: occhild.lo $(LOCAL_LIBS)
$(LINK_PROG) occhild.lo $(LOCAL_LIBS) $(ALL_LIBS)
Re: svn commit: r574113 - /apr/apr/trunk/test/Makefile.in
Posted by Bojan Smojver <bo...@rexursive.com>.
On Mon, 2007-09-10 at 14:53 +1000, Bojan Smojver wrote:
> The problem was that nobody ever knew that any of the tests actually
> failed, because make wouldn't stop.
http://issues.apache.org/bugzilla/show_bug.cgi?id=42725
--
Bojan
Re: svn commit: r574113 - /apr/apr/trunk/test/Makefile.in
Posted by Bojan Smojver <bo...@rexursive.com>.
On Mon, 2007-09-10 at 00:03 -0500, William A. Rowe, Jr. wrote:
> The full list of failures should be reported, otherwise
> you only discover the first flaw and have no clue of the scope of the issue.
OK, we can do that too, it's not difficult.
--
Bojan
Re: svn commit: r574113 - /apr/apr/trunk/test/Makefile.in
Posted by "William A. Rowe, Jr." <wr...@rowe-clan.net>.
Bojan Smojver wrote:
>
> The idea, as understand it, is that if you run make, things should run
> from beginning to end, unattended. If tests fail midway through the run,
> then obviously things aren't working and they need attention. Each
> tester can then get into test directory and run other tests from the
> sequence by hand, if desired/required.
Bloody heck no. The full list of failures should be reported, otherwise
you only discover the first flaw and have no clue of the scope of the issue.
(One test? Everything that touches a date type?)
Then we aught to bomb out and emit a non-0 result from make.
Re: svn commit: r574113 - /apr/apr/trunk/test/Makefile.in
Posted by Bojan Smojver <bo...@rexursive.com>.
On Sun, 2007-09-09 at 22:15 -0700, Sander Temme wrote:
> Make shouldn't stop, but run all the tests, quietly, and, if there
> are failures, emit a list of those failures and fail the target. At
> the end, not at the first failure.
Hopefully r574134/5 should be better. Any dramas, let me know.
--
Bojan
Re: svn commit: r574113 - /apr/apr/trunk/test/Makefile.in
Posted by Sander Temme <sc...@apache.org>.
On Sep 9, 2007, at 9:53 PM, Bojan Smojver wrote:
> On Sun, 2007-09-09 at 21:35 -0700, Sander Temme wrote:
>
>> Aaaah! What about the other tests later in the sequence? Usually
>> test targets complete every test, regardless of failures.
>
> The problem was that nobody ever knew that any of the tests actually
> failed, because make wouldn't stop.
Make shouldn't stop, but run all the tests, quietly, and, if there
are failures, emit a list of those failures and fail the target. At
the end, not at the first failure.
S.
--
Sander Temme
sctemme@apache.org
PGP FP: 51B4 8727 466A 0BC3 69F4 B7B8 B2BE BC40 1529 24AF
Re: svn commit: r574113 - /apr/apr/trunk/test/Makefile.in
Posted by Bojan Smojver <bo...@rexursive.com>.
On Sun, 2007-09-09 at 21:35 -0700, Sander Temme wrote:
> Aaaah! What about the other tests later in the sequence? Usually
> test targets complete every test, regardless of failures.
The problem was that nobody ever knew that any of the tests actually
failed, because make wouldn't stop.
The idea, as understand it, is that if you run make, things should run
from beginning to end, unattended. If tests fail midway through the run,
then obviously things aren't working and they need attention. Each
tester can then get into test directory and run other tests from the
sequence by hand, if desired/required.
--
Bojan
Re: svn commit: r574113 - /apr/apr/trunk/test/Makefile.in
Posted by Sander Temme <sc...@apache.org>.
On Sep 9, 2007, at 6:17 PM, bojan@apache.org wrote:
> Author: bojan
> Date: Sun Sep 9 18:17:48 2007
> New Revision: 574113
>
> URL: http://svn.apache.org/viewvc?rev=574113&view=rev
> Log:
> Any test returning non-zero should stop make
Aaaah! What about the other tests later in the sequence? Usually
test targets complete every test, regardless of failures.
S.
--
Sander Temme
sctemme@apache.org
PGP FP: 51B4 8727 466A 0BC3 69F4 B7B8 B2BE BC40 1529 24AF