Craig Tiller
|
8b976d0c24
|
Fixes
|
10 years ago |
|
Craig Tiller
|
3054756462
|
Fix in-flight cancellations
|
10 years ago |
Craig Tiller
|
4f9727326e
|
Fix memory usage
|
10 years ago |
|
Craig Tiller
|
de64862323
|
Remove ternary operators here
|
10 years ago |
|
Craig Tiller
|
b1d99bf097
|
Better fix for segfault
|
10 years ago |
|
Craig Tiller
|
ebf94bf558
|
Fix segfault
|
10 years ago |
|
Craig Tiller
|
37bbead483
|
Fix memory leaks
|
10 years ago |
|
Craig Tiller
|
ea3697bf4c
|
Fix memory leak
|
10 years ago |
David Klempner
|
e5437de181
|
Add a missing mdstr_unref
This fixes most of the asan reported leaks.
|
10 years ago |
|
Craig Tiller
|
e2b2b1fb67
|
Fix check for whether we should write to prevent infinite loop
|
10 years ago |
|
Craig Tiller
|
dcde31fad7
|
Fix check for whether we should write to prevent infinite loop
|
10 years ago |
|
Craig Tiller
|
7bd9b99d97
|
Autohint write buffering
If there's an operation already scheduled after the one we are about to start, hint that write buffering is desired.
|
10 years ago |
|
Craig Tiller
|
d6731628bc
|
Cleanup documentation
|
10 years ago |
|
Craig Tiller
|
d642dcf920
|
Use unsigned for shifts
|
10 years ago |
|
Craig Tiller
|
166e25002c
|
First test passes
|
10 years ago |
|
Craig Tiller
|
e5d683ca7e
|
Add some documentation to call.c
|
10 years ago |
|
Craig Tiller
|
fb189f826e
|
Flesh out batch API
|
10 years ago |
|
Craig Tiller
|
c12fee6a04
|
Simplify call.c
Reduce duplication of handling in many places and simplify internally
tracked state.
|
10 years ago |
|
Craig Tiller
|
fef766933e
|
Back-end for new core API
|
10 years ago |
|
Craig Tiller
|
daceea8670
|
Add some documentation, simplify state
|
10 years ago |
|
Craig Tiller
|
c18c56e40c
|
Split master and reqinfo for clearer code
|
10 years ago |
|
Craig Tiller
|
c4f0ebe63b
|
Prepare for the new batch call API.
Rename all core API functions that are on their way to deprecation with
an _old tag across all wrappings.
|
10 years ago |
|
Craig Tiller
|
566316f3a5
|
Split legacy code to end of file
|
10 years ago |
|
Craig Tiller
|
a7170ac4ad
|
Get the recv close edge right
|
10 years ago |
|
Craig Tiller
|
9cc6141f99
|
clean up finish a little more
|
10 years ago |
|
Craig Tiller
|
39af15104d
|
Cover all edges in the switch (not correctly yet)
|
10 years ago |
|
Craig Tiller
|
12af2bd560
|
Remove unnecessary assert
|
10 years ago |
|
Craig Tiller
|
a7cac78516
|
Prepare for the new batch call API.
Rename all core API functions that are on their way to deprecation with
an _old tag across all wrappings.
|
10 years ago |
|
Craig Tiller
|
6bc1f2761b
|
Fix array size
|
10 years ago |
|
Craig Tiller
|
8f4f6e207f
|
Check for null pointer
|
10 years ago |
|
Craig Tiller
|
7a9d79244f
|
Fix simple request
|
10 years ago |
|
Craig Tiller
|
9a3f2b2433
|
Look in the right array element
|
10 years ago |
|
Craig Tiller
|
1c1419011a
|
Get server status send working
|
10 years ago |
|
Craig Tiller
|
05140d032b
|
Made call.c compile again
|
10 years ago |
|
Craig Tiller
|
1e0d4c43d1
|
Rename & semantic fixup progress
|
10 years ago |
|
Craig Tiller
|
97c943db8d
|
Remove unused variable
|
10 years ago |
|
Craig Tiller
|
03d5bf7e94
|
Rename some stuff
|
10 years ago |
|
Craig Tiller
|
c689ca31e2
|
Free mdout buffers
|
10 years ago |
|
Craig Tiller
|
05b034dbd1
|
Free the messages
|
10 years ago |
|
Craig Tiller
|
1500b952ec
|
Free the messages
|
10 years ago |
|
Craig Tiller
|
68f55a0804
|
Free the messages
|
10 years ago |
|
Craig Tiller
|
3a4749fd79
|
Call owns all metadata until its destroyed
|
10 years ago |
|
Craig Tiller
|
79ce823fb1
|
Cleanup unused state
|
10 years ago |
|
Craig Tiller
|
cffbcb7be8
|
Actually, dont start reading until initial metadata send is done
|
10 years ago |
|
Craig Tiller
|
e6b92e62f1
|
Dont request more data until the first message is received
|
10 years ago |
|
Craig Tiller
|
f5c914694d
|
Fix spam
|
10 years ago |
|
Craig Tiller
|
50524cc67d
|
Properly deal with end of stream
|
10 years ago |
|
Craig Tiller
|
4acdbb3352
|
Signedness fix
|
10 years ago |
|
Craig Tiller
|
aef25da3e8
|
Allow asynchronous call destruction in some cases
If there's a chance we're being called via a callback, we need to delay destruction until later. Otherwise, for performance, destroy it inline.
|
10 years ago |
|
Craig Tiller
|
b58f95d285
|
Fix early edge cases
|
10 years ago |
