Bugzilla – Full Text Bug Listing
| Summary: | [dstore] Log_To_File no longer default value for log_location in rsecomm.properties | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Product: | [Tools] Target Management | Reporter: | Onno Van den Troost <onno> | ||||||
| Component: | RSE | Assignee: | David McKnight <dmcknigh> | ||||||
| Status: | RESOLVED FIXED | QA Contact: | Martin Oberhuber <mober.at+eclipse> | ||||||
| Severity: | normal | ||||||||
| Priority: | P3 | CC: | dmcknigh | ||||||
| Version: | 3.2.1 | ||||||||
| Target Milestone: | 3.4 M5 | ||||||||
| Hardware: | Other | ||||||||
| OS: | other | ||||||||
| Whiteboard: | |||||||||
| Bug Depends on: | |||||||||
| Bug Blocks: | 366438 | ||||||||
| Attachments: |
|
||||||||
|
Description
Onno Van den Troost
We think that [351993] [dstore] not able to connect to server if .eclipse folder not available is the root cause for the changed behavior. Onno, the default for the rsecomm.properties file in RSE 3.2.x is this: log_location=Log_To_File Do you not have that set by default? Created attachment 208186 [details]
test patch
I see a potential issue in the case where rsecomm.log has not yet been created although I'm not sure this is the issue you're hitting. Can you see if this patch makes any difference?
Alright, so now I understand that you guys don't actually use rsecomm.properties. This is something I wasn't aware of. Right now the ServerLogger falls back to use stdout if there is an exception finding the rsecomm.properties file. To preserve the original behaviour (for the case where there is no rsecomm.properties), I can change that back to use the file-based logging. Created attachment 208192 [details]
updated patch to fall back to log to file rather than log to stout
David, By default, when a user connects to the host, an existing rsecomm.log is wiped out and it starts again from scratch. We have a mechanism to rename rsecomm.log to rsecomm.last during logon, to preserve the last log for support purposes. I don't know enough about the internals, but it makes sense to me that this rename happens before RSE gets its hands on it, so there is indeed no rsecomm.log as far as RSE is concerned. Regarding rsecomm.properties. Things get messy in the single server setup when log_location is not set to Log_To_File, so we're shielding customers from shooting themselves in the foot by not providing the directive in the sample rsecomm.properties we provide. But we do have a rsecomm.properties file, used for the debug_level directive. I hope this info still fits in your current understanding and resolution of the problem. Onno (In reply to comment #6) > David, > > By default, when a user connects to the host, an existing rsecomm.log is wiped > out and it starts again from scratch. We have a mechanism to rename rsecomm.log > to rsecomm.last during logon, to preserve the last log for support purposes. > > I don't know enough about the internals, but it makes sense to me that this > rename happens before RSE gets its hands on it, so there is indeed no > rsecomm.log as far as RSE is concerned. > > Regarding rsecomm.properties. Things get messy in the single server setup when > log_location is not set to Log_To_File, so we're shielding customers from > shooting themselves in the foot by not providing the directive in the sample > rsecomm.properties we provide. But we do have a rsecomm.properties file, used > for the debug_level directive. > > I hope this info still fits in your current understanding and resolution of the > problem. > > Onno Onno, alright that helps explain things. Assuming the patch here solves the issue, do you need this backported for RSE 3.2.x? yes, please backport to 3.2.x I've committed the change. |