This is the mail archive of the systemtap@sourceware.org mailing list for the systemtap project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [patch] Use user specific systemtap cache in testsuite


Hi,

On Wed, 2008-05-28 at 13:29 +0200, Mark Wielaard wrote:
> make installcheck still depends on clean which does try to remove the
> cache dirs (but might silently fail). I don't think this is really
> necessary anymore.

Franks said on irc he also thought that was so. So I removed the
dependency. Frank also pointed out that I redirected stderr wrongly. And
that I forgot to check in the generated Makefile.in. All fixed as
follows:

2008-05-28  Mark Wielaard  <mwielaard@redhat.com>

    * testsuite/Makefile.am (clean-local): Correct redirect of stderr.
    (installcheck): Don't depend on clean.

Tested in various ways, with and without sudo. Seems to work fine and
give consistent test results for me now.

> But currently the testsuite hangs at the end because of this commit:
> 
> commit 7938bd829dfd82c40b35d536b5913e666c1454d2
> Author: Frank Ch. Eigler <fche@elastic.org>
> Date:   Tue May 27 17:40:00 2008 -0400
> 
>     testsuite: use catch{close}catch{wait} to ensure child process
> pipe/zombie cleanup
> 
> It seems not everybody is seeing this though. And I don't immediately
> see what is causing the hang in this patch.

We couldn't figure out what it really was, but Frank has reverted this
patch for now.

Cheers,

Mark
diff --git a/testsuite/Makefile.am b/testsuite/Makefile.am
index fe089f5..b66bb75 100644
--- a/testsuite/Makefile.am
+++ b/testsuite/Makefile.am
@@ -9,7 +9,7 @@ all-local:
 
 clean-local:
 	-rm -f ./stap site.exp systemtap.sum systemtap.log
-	-rm -rf .systemtap* .cache_test* 2>&1 /dev/null
+	-rm -rf .systemtap* .cache_test* 2>/dev/null
 
 DEJAZILLA=@dejazilla@
 
@@ -20,7 +20,7 @@ check-local:
 	if test -n "$(DEJAZILLA)"; then mail $(DEJAZILLA) < systemtap.sum; fi
 
 # but installcheck does not follow an implicit check-DEJAGNU, go figure
-installcheck: clean site.exp
+installcheck: site.exp
 	-$(MAKE) $(AM_MAKEFLAGS) check-DEJAGNU RUNTESTFLAGS="$(RUNTESTFLAGS) --tool_opts install"
 	if test -n "$(DEJAZILLA)"; then mail $(DEJAZILLA) < systemtap.sum; fi
 
diff --git a/testsuite/Makefile.in b/testsuite/Makefile.in
index cc76cad..c8b07bd 100644
--- a/testsuite/Makefile.in
+++ b/testsuite/Makefile.in
@@ -459,7 +459,7 @@ all-local:
 
 clean-local:
 	-rm -f ./stap site.exp systemtap.sum systemtap.log
-	-rm -rf .systemtap .cache_test 2>/dev/null
+	-rm -rf .systemtap* .cache_test* 2>/dev/null
 
 # automake's dejagnu library already runs check-DEJAGNU before check-local
 # That's why we need to add "execrc" to $(RUNTEST) - to ensure that this
@@ -468,7 +468,7 @@ check-local:
 	if test -n "$(DEJAZILLA)"; then mail $(DEJAZILLA) < systemtap.sum; fi
 
 # but installcheck does not follow an implicit check-DEJAGNU, go figure
-installcheck: clean site.exp
+installcheck: site.exp
 	-$(MAKE) $(AM_MAKEFLAGS) check-DEJAGNU RUNTESTFLAGS="$(RUNTESTFLAGS) --tool_opts install"
 	if test -n "$(DEJAZILLA)"; then mail $(DEJAZILLA) < systemtap.sum; fi
 # Tell versions [3.59,3.63) of GNU make to not export all variables.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]