#2993 Memory leak?

[Donal K. Fellows]

compile-12.1 fails for me (config: Linux, Fedora Core
2, Tcl HEAD built with --enable-symbols=all,
--enable-threads)

==== compile-12.1 testing literal leak on interp delete
FAILED
==== Contents of test case:

proc getbytes {} {
set lines [split [memory info] "\n"]
puts [lindex $lines 3]
lindex $lines 3 3
}

set end [getbytes]
for {set i 0} {$i < 5} {incr i} {
interp create foo
foo eval {
namespace eval bar {}
}
interp delete foo
set tmp $end
set end [getbytes]
}
rename getbytes {}
set leak [expr {$end - $tmp}]

---- Result was:
20736
---- Result should have been (exact matching):
0
==== compile-12.1 FAILED

[miguel sofer]

Logged In: YES
user_id=148712

Reproed under --enable-symbols=mem (no threads); leak is
5328 here

[miguel sofer]

Logged In: YES
user_id=148712

Leak is real, and independent ofcompile flags. With a normal
build, you can watch (in top) as a tclsh running
while i {interp create a; interp delete a}
grows in memory used.

[Jeffrey Hobbs]

Logged In: YES
user_id=72656

I cannot confirm this for 8.4.9 or HEAD on Windows. Unix
only?? Testing with:

make shell
% while 1 { interp delete [interp create a] }

[miguel sofer]

Logged In: YES
user_id=148712

The leak was not there in 8.5a1

[Don Porter]

Logged In: YES
user_id=80530

leak is new since 2004-11-15.

[Don Porter]

Logged In: YES
user_id=80530

leak is new since 2004-11-28.

[Don Porter]

Logged In: YES
user_id=80530

Ignore those comments.

Appears the failed test doesn't
show up on Solaris ?

[Don Porter]

Logged In: YES
user_id=80530

ok, on a better test platform,
confirmed the leak started
between 2004-11-30 and
2004-12-01. very likely a
problem in my commit.

[Don Porter]

Logged In: YES
user_id=80530

leak is rooted in the
post-script-evaluation
portion of Tcl_Init.

[Don Porter]

Logged In: YES
user_id=80530

leak happens in the call to
TclpGetEncodingNameFromEnvironment()

[Don Porter]

Logged In: YES
user_id=80530

leak happens in the
Tcl_GetEncoding() call
that returns NULL.

[Don Porter]

Logged In: YES
user_id=80530

Leak is in MakeFileMap()
in tclEncoding.c

root cause was confusion about the
refCounting of Tcl_Obj's returned
by TclPathPart().

Committing fix...

thanks to the reporters and testers.