Change parameter type #28
Conversation
| do { | ||
| ret = recv(fd, buf, (size_t)(size-offset), 0); | ||
| while(offset < size) { | ||
| ret = recv(fd, buf+offset, size-offset, 0); |
There was a problem hiding this comment.
That's actually a behaviour change no? I believe this is indeed a needed fix, but that kind of things shouldn't be buried in a commit that should have no functional changes.
There was a problem hiding this comment.
Or at least you should details how you refactored the loop in the commit message.
There was a problem hiding this comment.
There are two things. The refactoring (including the loop refactor) is one commit. Then there's the buf+offset, which is not part of the refactoring as far as I can see (but could be wrong). That's a second commit. But I could be wrong of course.
src/shim.c
Outdated
|
|
||
| if ( fd < 0 || ! buf ) { | ||
| if ( fd < 0 || ! buf) { |
There was a problem hiding this comment.
This is a spurious white space changes. The rule is no such thing sharing the same commit with other changes. The indentation in the shim is a bit all over the place, seems like we have hard time decided what should be used:
if(foo)
if (foo)
if ( foo )
and now:
if ( foo)
Life may well be too short for this, but opened #31 anyway. This isn't high priority.
|
Just need an answer to the buf+offset change and if indeed a bug fix, split it in a separate commit. |
|
yes, that indeed is a bug fix that I noticed while refactoring a bit, i'll split it up |
|
The point was that you could add a const to the first parameter of serialize_frame: But doesn't matter that much. |
|
lgtm
|
|
Sorry, lgtm'ed while there's still a change pending. |
amshinde
force-pushed
the
change-param-type
branch
from
e945706 to
be0135e
Compare
|
@dlespiau Changes pushed |
| @@ -391,11 +391,16 @@ read_frame(struct cc_shim *shim) | |||
| abort(); | |||
| } | |||
|
|
|||
| if (! read_wire_data(shim->proxy_sock_fd, buf, (ssize_t)size)) { | |||
There was a problem hiding this comment.
Was this line removal on purpose?
There was a problem hiding this comment.
shuffled between rewriting commits. fixed
Check that the header length received from proxy is at least the minimum length defined by the proxy protocol. We need to define an upper limit on this in the future. Signed-off-by: Archana Shinde <[email protected]>
In case of the incomplete reads, the data needs to be stored at the right place by incrementing the buffer address by offset, number of bytes read in the last operation. Signed-off-by: Archana Shinde <[email protected]>
Change the type of parameter "size" from size_t to ssize_t as this really needs to be an unsigned value. Change the loop to a while loop. Get rid of the casts in some of the calls to read_wire_data. Signed-off-by: Archana Shinde <[email protected]>
amshinde
force-pushed
the
change-param-type
branch
from
be0135e to
e1e5c05
Compare
No description provided.