tree: add missing div-by-zero error for float div #37774
Conversation
|
This change is |
There was a problem hiding this comment.
Reviewed 2 of 2 files at r1.
Reviewable status:complete! 0 of 0 LGTMs obtained (waiting on @jordanlewis and @mjibson)
pkg/sql/sem/tree/eval_test.go, line 255 at r1 (raw file):
{`1 % 0`, `zero modulus`}, {`1 / 0`, `division by zero`}, {`1::float / 0::float`, `division by zero`},
why does the test go here instead of in testdata/eval/float_decimal_div? is it because it's an error?
There was a problem hiding this comment.
Reviewable status:
pkg/sql/sem/tree/eval_test.go, line 255 at r1 (raw file):
Previously, rafiss (Rafi Shamim) wrote…
why does the test go here instead of in
testdata/eval/float_decimal_div? is it because it's an error?
Yes, as far as I can tell, the datadriven package that we use for the eval tests here doesn't know how to expect errors yet. That would be a nice feature addition - for now, this other test has a good way of specifying expected errors.
jordanlewis
force-pushed
the
fix-div-float-0
branch
from
060e03d
to
2f35aa4
Compare
There was a problem hiding this comment.
Also had to remove 2 older tests that expected float div-by-0 to return non-errors.
Reviewable status:
There was a problem hiding this comment.
Reviewable status:
jordanlewis
force-pushed
the
fix-div-float-0
branch
from
2f35aa4
to
0828313
Compare
Previously, dividing a float by 0 would return -Inf, instead of returning a divide by 0 error as expected. Release note (bug fix): dividing floats by zero now returns an error, instead of -Inf.
jordanlewis
force-pushed
the
fix-div-float-0
branch
from
0828313
to
9cdc5cf
Compare
|
bors r+ TFTRs! |
37774: tree: add missing div-by-zero error for float div r=jordanlewis a=jordanlewis Previously, dividing a float by 0 would return -Inf, instead of returning a divide by 0 error as expected. Release note (bug fix): dividing floats by zero now returns an error, instead of -Inf. Co-authored-by: Jordan Lewis <jordanthelewis@gmail.com>
Build succeeded |
Previously, dividing a float by 0 would return -Inf, instead of
returning a divide by 0 error as expected.
Release note (bug fix): dividing floats by zero now returns an error,
instead of -Inf.