On Mon, Feb 09, 2009 at 04:15:01PM +0100, Dr. Werner Fink wrote:
> On Fri, Feb 06, 2009 at 09:02:33AM -0500, Glenn Fowler wrote:
> >
> > thanks werner
> > I will be posting 2002-02-02 beta src later today
> >
> > On Fri, 6 Feb 2009 12:17:34 +0100 Dr. Werner Fink wrote:
> > > Hi,
> >
> > > this is the message which was found by a tester. The tar balls
> > > ast-ksh and INIT are from 2008-12-12. It seems that 2009-01-20
> > > has fixed that probem but the test suite shows:
> >
> > > test basic.sh begins at 2009-02-06+10:46:30
> > > /tmp/ksh2454x: line 2: 2556: (coredump)
> > > test basic.sh passed at 2009-02-06+10:46:43 [ 90 tests 0 errors ]
> >
> > > test coprocess.sh begins at 2009-02-06+10:46:52
> > > coprocess.sh: line 95: 2664:
> > > coprocess.sh: line 97: 2667:
> > > coprocess.sh[97]: parent coprocess prevents script coprocess
> > > coprocess.sh: line 1: 2709:
> > > test coprocess.sh failed at 2009-02-06+10:47:02 with exit code 1 [ 25
> > > tests 1 error ]
> >
> > > this was Version JM 93t+ 2009-01-20. Please note that the
> > > tests are done within an special environment without having
> > > a terminal around.
>
> It looks like a setjmp/setsigjmp -> longjmp/siglongjmp goes
> wrong here. This because it is not a real SIGSEGV but only
> the message about. The script sniplet which fails is a three
> liner:
>
> myfilter() { x=$(print ok | cat); print -r -- $SECONDS;}
> set -o pipefail
> sleep 3 | myfilter
>
> please note that if I add a line like
>
> type cat
>
> before the first line it works. To reproduce this I've
> to deinstall ksh and run the test within the build
> tree like:
>
> base=src/cmd/ksh93
> root=$(echo ${PWD}/arch/linux*)
> test=${PWD}/${base}/tests
> pushd ${test}
> LD_LIBRARY_PATH=${root}/lib ${root}/bin/ksh /tmp/scriptlet
>
> which is simply what is done within the build environment.
> Please note that I using shared libraries for libast, libcmd,
> libdll, and libshell. Therefore the LD_LIBRARY_PATH variable
> is set.
This change
--- src/cmd/ksh93/sh/xec.c
+++ src/cmd/ksh93/sh/xec.c
@@ -1440,7 +1452,12 @@
pid_t savepgid = job.curpgid;
job.curpgid = 0;
if(shp->subshell)
- sh_subtmpfile(1);
+ {
+ if(shp->subshare)
+ sh_subtmpfile(0);
+ else
+ sh_subfork();
+ }
shp->inpipe = pvo;
shp->outpipe = pvn;
pvo[1] = -1;
seems to cause the false coredump message ... using only
sh_subtmpfile(1) removes the messages but cause an other
message from shell.sh
subshell.sh[244]: $() of pipeline hangs
subshell.sh[245]: ${} sub of pipeline hangs
...
Nevertheless the random message
test coprocess.sh begins at 2009-02-10+14:51:33
coprocess.sh: line 97: 12715:
coprocess.sh[97]: parent coprocess prevents script coprocess
coprocess.sh: line 1: 12755:
still happens sometimes.
Werner
--
"Having a smoking section in a restaurant is like having
a peeing section in a swimming pool." -- Edward Burr
_______________________________________________
ast-developers mailing list
[email protected]
https://mailman.research.att.com/mailman/listinfo/ast-developers
