| Ticket UUID: | 1563538 | ||
| Title: | append+env bug | ||
| Type: | Bug | Created on: | 2006-09-22 14:25:46 |
| Submitter: | fridolin | Assigned to: | kennykb |
| Subsystem: | 07. Variables | Severity: | |
| Priority: | 5 Medium | Last modified: | 2006-10-09 16:14:48 |
| Status: | Open | Closed by: | |
| Resolution: | None | Closed on: | |
| Version: | obsolete: 8.5a4 | ||
| Description: | ||||
this bug could be seen on windows: I start with a command shell: c:\> set A=VALUE c:\> tclsh % append env(A) 123 VALUE123 % exit good! c:\> tclsh % puts $env(a) VALUE % append env(a) 123 VALUE123 % exit good! c:\> tclsh % # no puts % append env(a) 123 123 % puts $env(a) 123 % exit huh? | ||||
| User Comments: | ||||
msofer added on 2006-09-26 07:38:04:
Logged In: YES user_id=148712 I don't even a windows platform to test this, and it does seem windows specific. At least if there is no typo in the report, I infer that 1. windows env vars are not case sensitive 2. the case conversion is probably done by a read trace Note that we do *not* trigger read traces on append (even though we do on lappend). Comment in tclVar.c line 1604 is relevant, as well as the fact that TCL_TRACE_READS is not passed in line 2309 (HEAD refs). I do not see how we could fix this; maybe Kevin has an idea. fridolin added on 2006-10-09 16:14:48:
Logged In: YES user_id=13583 I don't know the internal details, but after the first access of env(a) it seems to work. Because the system environment shouldn't change after the start of an application, it could be solved by "initialize" env on startup (however this is to be done internally). This costs a startup penalty. (but maybe the read trace on env could be removed). | ||||
