diff options
| author | Daniel Martí <mvdan@mvdan.cc> | 2021-10-11 11:12:38 +0100 |
|---|---|---|
| committer | Daniel Martí <mvdan@mvdan.cc> | 2021-10-11 21:58:33 +0000 |
| commit | d973bb107e9142cf17e4a7f2666a71ed2d457e91 (patch) | |
| tree | 5ce64142a65ea2c3b8f634ec32788ee0e1fe84c4 /src/cmd/api | |
| parent | 70235351263caee1fd9840d1d652bf28778cd51d (diff) | |
| download | go-d973bb107e9142cf17e4a7f2666a71ed2d457e91.tar.xz | |
encoding/gob: follow documented io.EOF semantics
The docs say:
If the input is at EOF, Decode returns io.EOF and does not modify e.
However, the added test fails:
--- FAIL: TestDecodePartial (0.00s)
encoder_test.go:1263: 31/81: expected io.ErrUnexpectedEOF: EOF
encoder_test.go:1263: 51/81: expected io.ErrUnexpectedEOF: EOF
In particular, the decoder would return io.EOF after reading a valid
message for a type specification, and then hit EOF before reading a data
item message.
Fix that by only allowing a Decode call to return io.EOF if the reader
hits EOF immediately, without successfully reading any message.
Otherwise, hitting EOF is an ErrUnexpectedEOF, like in other cases.
Also fix a net/rpc test that, coincidentally, expected an io.EOF
as an error when feeding bad non-zero data to a gob decoder.
An io.ErrUnexpectedEOF is clearly better in that scenario.
Fixes #48905.
Change-Id: Ied6a0d8ac8377f89646319a18c0380c4f2b09b85
Reviewed-on: https://go-review.googlesource.com/c/go/+/354972
Trust: Daniel Martí <mvdan@mvdan.cc>
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Rob Pike <r@golang.org>
Diffstat (limited to 'src/cmd/api')
0 files changed, 0 insertions, 0 deletions
