Age | Commit message (Collapse) | Author |
|
In Persistent Group Chats (PGC), this will cause a use-after-free. This
test ensures that we fix this bug before merging PGC.
|
|
Currently, toxcore does not support UDP over proxies. In the future, we
can relax this by disabling UDP only if the proxy doesn't support it.
|
|
The C compiler warns because the value is initialised in a loop and used
outside of it. In this case, it's always initialised, but changing the
value of `NUM_PORTS` can change that.
|
|
|
|
I.e. make tox_new(0, 0) behave the same as tox_new(tox_options_new(0), 0).
Changing defaults in the options does not currently affect NULL options.
|
|
This can be used as a random stress test for toxcore.
Adjust the weights to make certain actions more or less likely.
|
|
Related: https://github.com/qTox/qTox/issues/5174
|
|
Better error messages, better sleep() call placements, etc.
Did not modify large chunk of function because I couldn't explain
it. Maybe I'll come back later once I've regained lost brain cells.
|
|
It was a list of #defines, but clearly this is a discrete and limited set
of named values, so an enum is more appropriate.
|
|
|
|
Tests all ports instead of just one, also adds some comments/changes
error messages to be more descriptive
|
|
This triggers a code path in Persistent Group Chats that causes a memory
leak. I'm adding this test now, so that we don't merge PGC without fixing
the memory leak first.
|
|
This test has never caught a bug. It's better to catch these with asan or
the likes.
|
|
`-fsanitize=address` also needs to be passed to the linker.
|
|
By making it run in a port range far away from other tests. This test
creates 100 DHTs and makes either itself or other tests run out of ports.
|
|
|
|
|
|
|
|
|
|
Signed overflow is undefined behaviour, so in debug mode, we want to make
it fail in a noisy way.
|
|
Fixes #900.
|
|
|
|
|
|
|
|
Fixes #868
Is made by:
1. Fix enumeration mechanism of Crypto_Connection instances in function send_crypto_packets()
(this item solves the main problem);
2. Disruption of correlation between parameters of different dimensions
(the parameters MAX_NUM_SENDPACKET_TRIES and UDP_DIRECT_TIMEOUT have different dimensions).
|
|
TODO: need a meanful log output
|
|
The Family stuff in toxcore is a big mess. I'm sure I saw a bunch of bugs
on the way, but I'm not verifying that code now, so the bugs stay.
|
|
These should register a proper logging callback so the messages don't go
devnull, but this at least ensures a logger is available.
|
|
|
|
Introduced by 643eea60bb9dcf4ecb33d64666b1bc77cbfd7438
|
|
|
|
|
|
Taken from:
https://github.com/TokTok/c-toxcore/commit/6b97acb773622f9abca5ef305cd55bdef1ecc484
|
|
Windows doesn't report network errors though errno, it has its own facilities.
|
|
|
|
|
|
|
|
|
|
|
|
We control the random functions in crypto_core, so we can make them
deterministic more easily. This will help test reproducibility in the
future.
|
|
|
|
|
|
|
|
Makes debugging easier.
|
|
This only works with clang/llvm. Coveralls support needs to be added later.
|
|
The new clang version makes gcov segfault. Also, coveralls needs another
package that it doesn't install the right version of by default.
|
|
This is useful for debugging a function that doesn't have a logger
available. It should not be used in production code, since it outputs to
stderr.
|
|
|
|
It's not supported in mingw.
See https://github.com/TokTok/c-toxcore/issues/786.
|
|
|