| Ticket UUID: | a366c6efee1f70cc9b27f949d41afe3ee77c0f08 | ||
| Title: | [ledit] corrupts string literal value | ||
| Type: | Bug | Created on: | 2023-07-14 14:59:26 |
| Submitter: | dgp | Assigned to: | apnadkarni |
| Subsystem: | - New Builtin Commands | Severity: | Severe |
| Priority: | 5 Medium | Last modified: | 2023-07-16 07:27:34 |
| Status: | Closed | Closed by: | apnadkarni |
| Resolution: | Fixed | Closed on: | 2023-07-16 07:27:34 |
| Version: | trunk | ||
| Description: | ||||
The tests ledit-4.2 and socket_inet-13.1 interact and cause failure.
The interaction can be summarized by this demo script:
% proc demo {} {
set l { }
puts [list [ledit l 1 1] $l]
set result foo
append result " " bar
}
% demo
{} {}
foobar
The result "foobar" is wrong. It should be "foo bar".
It appears that the string literal " " has had its string value
transformed to "" before the append. This might be the result
of not managing value-sharing properly when the [ledit] command
changes the value of $l from { } to {}.
Hope that's enough to lead the [ledit] keepers to a fix.
| ||||
| User Comments: | ||||
jan.nijtmans added on 2023-07-14 15:15:38:
This smells like the TclDuplicatePureObj() danger somewhere. apnadkarni added on 2023-07-15 04:06:30:
Will take a look later today. Strangely enough, 8.7 which pretty much shares the ledit code does not seem to have this problem. Not that means much... Thanks Don for a very concise demo. apnadkarni added on 2023-07-15 04:40:36:
Found the issue, as always not where I suspected. Fix forthcoming. apnadkarni added on 2023-07-16 07:27:34:
Fixed in [fe628d36ce]. As to why it did not manifest in 8.7 is that the more sophisticated TclCheckEmptyString check in 9.0 triggered the bug. | ||||
