eval: remove wrapping from builtin errors #147195
Conversation
Release note: None
![blathers[crl]](https://avatars.githubusercontent.com/in/59700?s=40&u=95b789b1d594ac6b74f377aa7555d8d98d63240d&v=4){kind=avatar}
|
Your pull request contains more than 1000 changes. It is strongly encouraged to split big PRs into smaller chunks. 🦉 Hoot! I am a Blathers, a bot for CockroachDB. My owner is dev-inf. |
|
This change is |
yuzefovich
force-pushed
the
pg_regress_builtins
branch
from
db6f19f to
bf02be1
Compare
yuzefovich
requested review from
kyle-a-wong,
arjunmahishi,
aa-joshi,
Abhinav1299,
herkolategan,
golgeek,
kev-cao and
rytaft
and removed request for
a team
yuzefovich
requested review from
a team,
kev-cao,
mgartner and
DrewKimball
and removed request for
kyle-a-wong,
arjunmahishi,
aa-joshi,
Abhinav1299,
herkolategan,
golgeek,
kev-cao,
rytaft,
a team and
mgartner
|
Commits 2 through 5 will be squashed into a single one, but I kept them separate for ease of reviewing. I looked into this because I left a TODO for myself to remove this wrapping context. That said, I don't feel too strongly about doing this, so if people find the current behavior preferable, I'm fine not merging this PR and simply removing the TODO. |
There was a problem hiding this comment.
Reviewed 4 of 4 files at r1, 3 of 3 files at r2, 46 of 46 files at r3, 2 of 2 files at r4, 31 of 31 files at r5, 8 of 8 files at r6, all commit messages.
Reviewable status:complete! 1 of 0 LGTMs obtained (waiting on @yuzefovich)
This commit removes the code to perform error wrapping around the errors produced by the builtin function which provides a bit more context (like the function name). This wrapping dates all the way back to a commit from 2015 716837e. Over the years it was expanded to also provide some more context to the sentry reports (7b703de). I don't think either of these provide much value since it's usually clear which builtin the error corresponds to and having just the builtin name in the sentry report doesn't make it any more actionable. Removing of this logic allows us to match errors as they are produced by Postgres. Release note: None
Release note: None
yuzefovich
force-pushed
the
pg_regress_builtins
branch
from
bf02be1 to
93bd46c
Compare
|
TFTR! bors r+ |
|
Build succeeded: |
roachtest/pg_regress: accept recent diff
Fixes: #147252.
eval: remove wrapping from builtin errors
This PR removes the code to perform error wrapping around the errors produced by the builtin function which provides a bit more context (like the function name). This wrapping dates all the way back to a commit from 2015 716837e. Over the years it was expanded to also provide some more context to the sentry reports (7b703de). I don't think either of these provide much value since it's usually clear which builtin the error corresponds to and having just the builtin name in the sentry report doesn't make it any more actionable. Removing of this logic allows us to match errors as they are produced by Postgres.
builins: unify an error message in substring with PG
Epic: None
Release note: None