look at the body to determine if it has content or not #149
Conversation
Signed-off-by: Ivan Porto Carrero <ivan@flanders.co.nz>
Codecov Report
@@ Coverage Diff @@
## master #149 +/- ##
==========================================
- Coverage 80.9% 80.59% -0.32%
==========================================
Files 40 40
Lines 2299 2329 +30
==========================================
+ Hits 1860 1877 +17
- Misses 337 347 +10
- Partials 102 105 +3
Continue to review full report at Codecov.
|
| if p.underlying.Buffered() > 0 { | ||
| return true | ||
| } | ||
| b, err := p.underlying.Peek(1) |
There was a problem hiding this comment.
Will this return pre-decoded bytes? If so, we may need to be careful about the the fact that an empty entity-body in chunked transfer-encoding is represented as a single ascii "0" after headers, and before the terminating CRLF.
Something like:
..headers...
transfer-encoding: chunked\r\n
\r\n
0\r\n
If this underlying was given bytes that already handles the transfer "decoding" then we should be safe here, but I can't tell exactly from the surrounding code just yet.
There was a problem hiding this comment.
The Go documentation is slightly comforting, but not specific:
178 // For server requests, the Request Body is always non-nil
179 // but will return EOF immediately when no body is present.
180 // The Server will close the request body. The ServeHTTP
181 // Handler does not need to.
182 Body io.ReadCloser
There was a problem hiding this comment.
Great: golang/go#11675 (comment) sounds like it handles transfer encodings, or at least it did 4 years ago.
| return false | ||
| } | ||
|
|
||
| rdr := newPeekingReader(r.Body) |
There was a problem hiding this comment.
my concern below is invalid if r.Body represents an entity-body after all transfer-encodings have been applied. but if it does not, then I think the peek(1) will observe ascii "0" (end of chunk marker) and mistakenly assume there is a body.
There was a problem hiding this comment.
the go stdlib takes care of this part
|
I have run a test with the patch and I can confirm it has resolved the issue we were seeing. Thank you! |
Add shim design doc.
fixes #148
Signed-off-by: Ivan Porto Carrero ivan@flanders.co.nz