Problems starting basic example #17
Comments
|
Huh, that is very weird indeed. What platform are you on? |
|
Also, can you try the |
|
The noria output doesn't add much: I'll build the griddle branch and get back.. |
|
Ah, sorry, yes, you'll also want to swap the |
|
Not much more info: It's on a macbook, rust version: frank@MacBook:~/git/noria (griddle)$ rustc --version
rustc 1.47.0-nightly (7e6d6e5f5 2020-08-16) |
|
That's really strange indeed. I haven't tested Noria much on macOS, as all the benchmarks and dev work I've done has been on Linux. If it works on macOS, it's by accident 😅 It certainly should not be segfaulting though. If you could figure out where it segfaults by running through |
|
Ok, first tried (without much success) to debug on Mac. Then, just to check, dusted off my linux box, and I have the exact same problem. I'll attempt debugging on the Linux box later, but for now I can say it doesn't seem to be Mac related. |
|
@ms705 Is this something you have seen before? |
|
Ok, I got the same issue now with gdb on my linux box, looks something like this: Does this clarify anything? |
|
Oh, that is very weird. That should only segfault if you have called the hidden |
|
I've tried it on another box. Nothing rust related on it. Rustup-ed, checked out all, did the same steps, with the same result. |
|
So, just to check: $ git clone https://github.com/mit-pdos/noria.git noria
$ git clone https://github.com/mit-pdos/noria-mysql.git noria-mysql
$ cd noria && cargo update
$ cd noria-mysql && cargo updateAnd then you run Can you try patching this function: in each match arm so that it prints whether the received thing is "local" or not? You'll probably have to modify the |
|
Ok, deleted all the Noria things, re-build & ran noria (same result). Pointed the noria-mysql dependencies to the local: (same result) Patched the proto.rs: Results in: And we see local! |
|
Thank you for digging! That's bizarre indeed...
and is only set to and that is only ever called in the Looking at the implementation, this line looks a little questionable, but I also suspect that it probably isn't hit anyway (maybe try adding a print): Just for my own sanity's sake, could you try removing the |
|
Ok, went back to this one. I've added some debug around the deserialization code: println!("Before deserialize!");
let result = usize::deserialize(deserializer).map(|p| LocalBypass(p as *mut T));
println!("After deserialize!");And it does get hit: Removing the tokio patch does not lead to discernible effect. I had a theory that it was somehow related to my mysql client (because that was the same accross all my attempts, an up-to-date mysql 8.0.21 client, but somehow an older client 8.0.15 does not connect at all. I also tried a JS based mysql client but that also does not seem to be able to connect at all. |
|
When cloning the latest versions of noria and noria-mysql (with noria-mysql patched to look locally), I got this same issue. I also saw this warning in the Noria-mysql build: I was able to fix this with the following Cargo.toml patch: along with a patch in noria-mysql to not refer to a function that has been removed: With this, I'm able to run those example SQL statements without anything crashing |
|
Yes, works for me! So I guess it would make sense then to release Noria 0.7, as the master of Noria MySQL depends on it...? |
Hi, I'm trying to get Noria working with JDBC. I'm aware of JDBC issues, I thought I'd take a stab at seeing where it fails, so I looked at issue #15 but I fail to even get the basics working.
Hopefully I'm doing something wrong.
I'm starting a zookeeper instance:
Seems alright. In a new shell, clean noria checkout:
Seems ok, in another shell, in a clean noria-mysql checkout:
Starts up fine.
Then in a final terminal:
Seems to work:
Create a table (like in #15):
But the next statement fails:
noria-mysql trace:
Noria terminates with a segfault:
Ideas?
The text was updated successfully, but these errors were encountered: