*Synopsis*: ksh93 hang in mysql build in SFW
Due to a change requested by <User 1-16P5I3>,
<User 1-16P5I3> is now the responsible engineer for:
CR 6859444 changed on Sep 22 2009 by <User 1-16P5I3>
=== Field ============ === New Value ============= === Old Value =============
Responsible Engineer <User 1-16P5I3>
See Also 5070274
====================== =========================== ===========================
*Change Request ID*: 6859444
*Synopsis*: ksh93 hang in mysql build in SFW
Product: solaris
Category: shell
Subcategory: korn93
Type: Defect
Subtype:
Status: 3-Accepted
Substatus:
Priority: 2-High
Introduced In Release:
Introduced In Build:
Responsible Engineer: <User 1-16P5I3>
Keywords:
=== *Description* ============================================================
A Makefile in the mysql 5.1 source in the SFW consolidation contains this code:
for f in `(cd $(srcdir); find suite -type f | egrep -v
'SCCS|row_lock')`; \
do \
d=$(DESTDIR)$(testdir)/`dirname $$f`; \
mkdir -p $$d ; \
$(INSTALL_DATA) $(srcdir)/$$f $$d ; \
done
The build hangs indefinitely at this point. The process tree contains a shell
with two children -- the find and the egrep. I can reproduce this with the
following shortened construct:
for f in `(find /var/sadm/pkg -type f)`; do :; done
Run on the commandline of any shell with "sh -c '<command>'", or substitute
"bash" or "zsh" or "/usr/has/bin/sh" for "sh" to see that it works in other
shells.
Note that the subshell is required, as is some amount of either IO or time
inside that subshell. I've tried saving the output of the command into a file
and simply catting that file inside the subshell, but then the command
completes successfully. I have also found that limiting the amount of output
from the find may allow the command to complete, too, but it doesn't appear to
be a strict issue of number of bytes. If I use head to limit the output, I can
zero in on a certain number of lines where the hang starts, but that will vary
from machine to machine, from one time of day to another (perhaps due to
general machine load), and will even differ between /usr/bin/head and
/usr/gnu/bin/head (the latter lets more lines through before the hang occurs).
*** (#1 of 1): 2009-07-12 01:03:06 GMT+00:00 <User 1-5Q-2233>
=== *Public Comments* ========================================================
See also the mailing list discussion starting at:
http://mail.opensolaris.org/pipermail/sfwnv-discuss/2009-July/004768.html
*** (#1 of 1): 2009-07-12 03:17:49 GMT+00:00 <User 1-5Q-1267>
=== *Workaround* =============================================================
Set SHELL to /usr/bin/bash for the target which calls this code:
--- usr/src/cmd/mysql-5-1/Makefile.sfw (SCCS v.1.2)
+++ usr/src/cmd/mysql-5-1/Makefile.sfw
@@ -158,6 +158,7 @@
$(GTAR) xjf $(MYSQL_DIR).tar.bz2
gpatch $(MYSQL_DIR)/storage/innobase/include/univ.i -i
patches/inline.patch
gpatch $(MYSQL_DIR)/scripts/mysql_config.sh -i
patches/mysql_config.patch
+ gpatch $(MYSQL_DIR)/mysql-test/Makefile.in -i patches/ksh-hang.patch
touch $(MYSQL_DIR)/configure
find $(MYSQL_DIR) -type d -exec /usr/bin/chmod 755 "{}" \;
find $(MYSQL_DIR) -type f -exec /usr/bin/chmod ugo+r "{}" \;
@@ -171,6 +172,7 @@
mv tmp/$(MYSQL_DIR) $(MYSQL_DIR_64); rm -rf tmp
gpatch $(MYSQL_DIR_64)/storage/innobase/include/univ.i -i
patches/inline.patch
gpatch $(MYSQL_DIR_64)/scripts/mysql_config.sh -i
patches/mysql_config.patch
+ gpatch $(MYSQL_DIR_64)/mysql-test/Makefile.in -i patches/ksh-hang.patch
touch $(MYSQL_DIR_64)/configure
find $(MYSQL_DIR_64) -type d -exec /usr/bin/chmod 755 "{}" \;
find $(MYSQL_DIR_64) -type f -exec /usr/bin/chmod ugo+r "{}" \;
--- /dev/null 2009-07-11 10:17:10.000000000 -0700
+++ usr/src/cmd/mysql-5-1/patches/ksh-hang.patch 2009-07-11
10:01:54.245425259 -0700
@@ -0,0 +1,11 @@
+--- Makefile.in.orig Mon Feb 2 08:37:22 2009
++++ Makefile.in Sat Jul 11 09:58:53 2009
+@@ -781,6 +781,8 @@
+ $(INSTALL_DATA) $(srcdir)/lib/My/*.pm $(distdir)/lib/My
+ -rm -rf `find $(distdir)/suite -type d -name SCCS`
$(distdir)/suite/row_lock
+
++install-data-local: SHELL=/usr/bin/bash
++
+ install-data-local:
+ $(mkinstalldirs) \
+ $(DESTDIR)$(testdir)/t \
*** (#1 of 2): 2009-07-12 01:03:06 GMT+00:00 <User 1-5Q-2233>
Better workaround: use $(...) instead of `(...)`.
Hangs:
for f in `(find /var/sadm/pkg -type f)`; do :; done
Doesn't hang:
for f in $(find /var/sadm/pkg -type f); do :; done
(This would explain why I've never noticed this bug, as I always use $() in
preference to ``.)
*** (#2 of 2): 2009-08-31 18:00:17 GMT+00:00 <User 1-5Q-10962>
=== *Additional Details* =====================================================
Targeted Release:
Commit To Fix In Build:
Fixed In Build:
Integrated In Build:
Verified In Build:
See Also: 6817601, 6859442, 6874461, 6876768, 5070274
Duplicate of:
Hooks:
Hook1:
Hook2:
Hook3:
Hook4:
Hook5:
Hook6:
Program Management:
Root Cause:
Fix Affects Documentation: No
Fix Affects Localization: No
=== *History* ================================================================
Date Submitted: 2009-07-12 01:03:06 GMT+00:00
Submitted By: <User 1-5Q-2233>
Status Changed Date Updated Updated By
3-Accepted 2009-08-31 09:29:40 GMT+00:00 <User 1-1SURPB>
=== *Service Request* ========================================================
Impact: Significant
Functionality: Primary
Severity: 2
Product Name: solaris
Product Release: osol_2009.06
Product Build:
Operating System: solaris
Hardware: generic
Submitted Date: 2009-07-12 01:03:06 GMT+00:00
=== *Multiple Release (MR) Cluster* - 0 ======================================
