OriginalBugID: 5099 Bug
Version: 8.3
SubmitDate: '2000-04-17'
LastModified: '2000-10-25'
Severity: CRIT
Status: UnAssn
Submitter: techsupp
OS: Digital Unix
OSVersion: Digital Unix 4.0E
Machine: AlphaStation 500
FixedDate: '2000-10-25'
ClosedDate: '2000-10-25'
Name:
Dr.James Reus
Extensions:
none
CustomShell:
none
ReproducibleScript:
# make test
:
pwd.test
reg.test
!!! bypassed 3 tests in reg-9, `reg brackets'
!!! bypassed 22 tests in reg-22, `reg multicharacter collating elements'
!!! bypassed 1 tests in reg-29, `reg incomplete matches'
regexp.test
registry.test
:
plit.test
stack.test
Stack overflow: pid 21180, proc tcltest, addr 0x11fdffb18, pc 0x12005f460
sh: 21180 Memory fault - core dumped
*** Exit 139
Stop.
ObservedBehavior:
core dump
DesiredBehavior:
completionn of all tests
Yuck. Why on earth is Tcl running out of stack space? This should not happen! (Are you running with a particularly small stack size ulimit set?)
This is a known problem when the stacksize limit is small. The stack.test for 8.4a2 has been improved to work around this, but we still need a generic stack space checker in Tcl. Alternatively (or additionally), we need to allow for setting the recursion limit from Tcl (using the C Tcl_SetRecursionLimit API), because that's what blows the stack. Each recursion level takes a bit of stack. Again alternatively, we can work on using less stack. However, that makes things more expensive in general because it causes more usage of runtime heap allocation.
See also bug#119353
Forget that! It was a completely erroneous comment. Sometimes I feel so dumb...
Logged In: YES
user_id=72656
The tests have been changed to be more aware of small stack
sizes with threaded builds.