-
Notifications
You must be signed in to change notification settings - Fork 12.9k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Run valgrind on a bot #13217
Comments
Nominating because I will be super embarrassed if we release 1.0 with memory leaks. |
P-high, 1.0 |
We are actually running valgrind on the snapshot bots: the basic problem here is that we have a broad suppression for most of memcheck because of #5856. |
I actually just turned valgrind off on all the bots. I recently created some new snapshot bots, and those have a very recent valgrind installed, and they uncovered a huge number of errors, most of which I don't think are relevant. I uncovered a few errors here and there, but the volume was too high to deal with and we need a new snapshot, so valgrind is actually turned off now. |
@alexcrichton given that this is a blocker for 1.0, I'd like to know what our plans are for resolving this. |
With the addition of |
Thinking back on this, I"m going to close this as fixed as we can selectively add tests to |
Check exit status of subcommands spawned by rustc_tools_util The git commands `git rev-parse --short HEAD` and `git log -1 --date=short --pretty=format:%cd` that clippy runs from its build script might fail with **"fatal: not a git repository (or any of the parent directories): .git"** if clippy is being built from a source tarball rather than a git repository. That message is written by git to stderr, and nothing is written to stdout. For `clippy-driver --version` this PR wouldn't make a difference because it treats empty stdout and failed spawns (`git` is not installed) identically: https://github.com/rust-lang/rust-clippy/blob/7ac242c3d0d3ee867a6c9cdcbdec986c408ac36f/rustc_tools_util/src/lib.rs#L35-L42 But other users of `rustc_tools_util` should be able to expect that the distinction between Some and None is meaningful. They shouldn't need extra code to handle None vs Some-and-empty vs Some-and-nonempty. --- changelog: none
We are not running valgrind anywhere right now, even though the linux dist builder says it's enabling valgrind, valgrind is not installed on that machine.
The makefiles probably should not silently disable valgrind like they appear to be doing.
The text was updated successfully, but these errors were encountered: