Fix PGCopyOutputStream out of order writes #1854
Conversation
Codecov Report
@@ Coverage Diff @@
## master #1854 +/- ##
============================================
+ Coverage 69.26% 69.29% +0.02%
- Complexity 4206 4208 +2
============================================
Files 197 197
Lines 17981 17983 +2
Branches 2914 2914
============================================
+ Hits 12455 12461 +6
+ Misses 4181 4175 -6
- Partials 1345 1347 +2 |
|
@davecramer This is ready too. Let's get this in for the upcoming release as it's a minor standalone fix. |
vlsi
force-pushed
the
master
branch
2 times, most recently
from
4c39f96
to
866c6a9
Compare
Codecov Report
@@ Coverage Diff @@
## release/42.2 #1854 +/- ##
==================================================
+ Coverage 69.21% 69.29% +0.07%
- Complexity 4193 4208 +15
==================================================
Files 197 197
Lines 17996 17983 -13
Branches 2916 2914 -2
==================================================
+ Hits 12456 12461 +5
+ Misses 4188 4175 -13
+ Partials 1352 1347 -5 |
|
anyone know why this just showed up again? @sehrope we should get this in. can you rebase and I will take care of it |
Adds a helper method to execute a query and reads a single text column as a string.
Adds a failing test that writes to a PGCopyOutputStream using both the write(byte[]) and write(ByteStreamWriter) APIs to catch internal buffering breaking the order.
…StreamWriter Fixes out of order writes when interleaving writes using buffered write(byte[]) and writeToCopy(ByteStreamWriter) APIs.
|
@davecramer No idea. I tried rebasing it locally to bring it up to date but am getting a bunch of unrelated conflicts. Was part of history on master rewritten? I cherry-picked the changes on to a v2 branch. Once it get's through travis (https://travis-ci.org/github/sehrope/pgjdbc/builds/742441641) I'll force push this original PR so it get's picked up. The individual commits are all localized so shouldn't be any true conflicts. |
|
History was not rewritten, however we did move some tags. Not sure why there are conflicts |
|
No I think something in master was definitely rewritten. I ran Not sure what happened but the same commits were published to the mailing list twice as arriving on master (linked below). Older commits in this PR but not in masterFix header on existing Parameter Status doc page, and include entry in the index html page. (#1845)commit 0fe766b fix: preserve unquoted unicode whitespace in array literals (#1266)commit 8ad48a6 remove postgresql-jre6 and postgresql-jre7 projects (#1848)commit c80add9 (fix-copy-out-stream, deprecate-legacy-pg-streams) Same commits (by substance) in master:Fix header on existing Parameter Status doc page, and include entry in the index html page. (#1845)commit 9962307 fix: preserve unquoted unicode whitespace in array literals (#1266)commit e8923c7 remove postgresql-jre6 and postgresql-jre7 projects (#1848)commit 866c6a9 |
|
@davecramer Looks like the cherry-picked branch passed Travis. Let me know if you want me to force it atop this or submit it as a separate PR to keep these commits in place (to track down the history rewrite). |
|
just force on top of it. Lets move forward |
sehrope
force-pushed
the
fix-pgcopyoutput-stream-out-of-order
branch
from
a17a24b
to
8395a50
Compare
|
so now that I think about it we did reorder some commits when we released a while ago. So I presume this causes a rewrite of history? |
|
Yes that must have been it. That's a big no-no and the master branch should be locked to prevent force-pushes. Can you do that from the GitHub console to prevent this in the future? It's fine to force-push PR and even pre-release branches if it's before they're published. But rewriting master or released branches causes all kinds of issues. |
|
There were extenuating circumstances.. |
This PR fixes #1777
It's just a fix for the out of order writes and an additional test for the corrected functionality. It does not deprecate the existing copy classes or add new replacements. We can follow up with that in a subsequent release.