refactor(c/driver_manager): differentiate errors from driver manager #1662
Conversation
lidavidm
force-pushed
the
gh-1653
branch
3 times, most recently
from
bbff6c1
to
e6387b9
Compare
| if (!error) return; | ||
| if (error->release && error->release != ReleaseError) { |
There was a problem hiding this comment.
Is it important to be able to reuse an existing error here that was previously initialized by the driver manager? It seems easier to explain to just always initialize a new one (or maybe it is worth explaining in a comment here?)
There was a problem hiding this comment.
I was preserving the behavior of just appending to an existing error (which admittedly is confusing - I can just get rid of that!)
| @@ -696,6 +749,7 @@ AdbcStatusCode AdbcDatabaseGetOption(struct AdbcDatabase* database, const char* | |||
| } else { | |||
| const auto it = args->options.find(key); | |||
| if (it == args->options.end()) { | |||
| SetError(error, "Option not found"); | |||
There was a problem hiding this comment.
Is is easy to pass on here which option was not found?
| @@ -948,6 +1006,7 @@ AdbcStatusCode AdbcDatabaseRelease(struct AdbcDatabase* database, | |||
| AdbcStatusCode AdbcConnectionCancel(struct AdbcConnection* connection, | |||
| struct AdbcError* error) { | |||
| if (!connection->private_driver) { | |||
| SetError(error, "AdbcConnectionCancel: must AdbcConnectionNew first"); | |||
There was a problem hiding this comment.
| SetError(error, "AdbcConnectionCancel: must AdbcConnectionNew first"); | |
| SetError(error, "AdbcConnectionCancel: must call AdbcConnectionNew first"); |
(...and a few below)
There was a problem hiding this comment.
I wonder if this diff would be suppressed if it were marked as linguist-generated in .gitattributes (it looks like it is currently linguist-vendored).
There was a problem hiding this comment.
I'll tweak it (though I think it wouldn't take effect until the next PR from what I recall seeing of how GitHub behaves)
Fixes #1653.