fix(server): Fix replication bug, add gdb opt to pytest #513
Conversation
| args = [self.params.path, *arglist] | ||
| if self.params.gdb: | ||
| args = ["gdb", "--ex", "r", "--args"] + args |
There was a problem hiding this comment.
I'm not sure this option is worth it.
Its helpful for me, because in case of a bad signal, the whole process stops. Once brought back to foreground, gdb is directly active and can be used to get stacktrace/dump variables/stack.
However, it doesn't allow customizing anything before start. Like selecting only a few instances, or setting breakpoints. The latter can be fixed by making the gdb parameter a string and forwarding it to gdb, so just --gdb becomes --gdb="break dflycmd.cc:100; break replica.cc:100"
dranikpg
force-pushed
the
fix-replica-bug
branch
from
af5fda8 to
6c31c69
Compare
Signed-off-by: Vladislav Oleshko <[email protected]>
dranikpg
force-pushed
the
fix-replica-bug
branch
from
6c31c69 to
9df4ee7
Compare
|
|
||
| // Start rdb saving. | ||
| SaveMode mode = shard == nullptr ? SaveMode::SUMMARY : SaveMode::SINGLE_SHARD; | ||
| std::error_code local_ec = snapshot->Start(mode, abs_path.generic_string(), scripts); | ||
| error_code local_ec = snapshot->Start(mode, full_path.generic_string(), scripts); |
There was a problem hiding this comment.
was it a bug, @dranikpg ?
There was a problem hiding this comment.
This is refactoring included. I broke it in the last PR unknowingly
|
LGTM. For the protocol, can you tell what the bugs were? |
|
|
The third change was increasing some general wait points in pytests, because the CI env is really slow. However, this has also the benefit of catching multi-threading bugs that are hard to reproduce on fast machines |
Fixes bug in replica.cc
Add option to run instances inside gdb