Fix: 2625 v3 - IDL only with extra mdsconnect and revert part of #2620#2632
Fix: 2625 v3 - IDL only with extra mdsconnect and revert part of #2620#2632mwinkel-dev wants to merge 4 commits intoMDSplus:alphafrom
Conversation
|
Hi @WhoBrokeTheBuild -- Before approving this PR for Fix 3, we should also discuss Fix 4. |
|
Python3 API is still unable to retrieve data in this branch / PR. See: #2625 |
|
Sean, we're still unable to replicate the python API bug on this branch, but we'll strip out the fix in Connections.c so that this PR is limited to just IDL changes, just in case. |
|
We have evaluated Fix 3 and Fix 4. Because Fix 3 has a smaller footprint (fewer files changed), we have decided to give priority to Fix 3. |
|
Update, we're now able to reproduce the python issue, it seems to be related to PR #2620 |
|
We are reverting the change to |
|
Reverting the |
WhoBrokeTheBuild
changed the title
|
I tested the changes again, and a set_database followed by a mdsconnect causes the proxied rdb connection to fail. See: #2625 |
|
Hi @sflanagan -- Please see my post in Issue #2625 (4-Oct-2023 ~8:55 pm EDT). My recollection (perhaps wrong) is that the broken database proxy is caused by a quirk in IDL itself. I will post test results later this evening that illustrate IDL's odd behavior. Correction: In addition to the IDL quirk, there is indeed an interaction between |
|
Added a partial fix for the broken database proxy. It is partial because it works a few times, but eventually fails. See Issue #2625 for details. |
|
Closing this PR, because the fix is not robust. This PR was never merged, thus no need to revert. For more details, refer to the comments in Issue #2625. |
3 participants
Also includes a change to Connections.c to do an integrity check of data structures when adding a connection (aka socket).