Show the memory usage and time to build a crate

[jyn514]

For crates which time out due to resource limits, there is very little indication of why: the logs show only 'process exited with signal SIGKILL'. It would be ideal to say why the process was killed and how much time and memory were used.

I wonder if docs.rs can use this same trick with /usr/bin/time -v?

$ cargo clean
$ /usr/bin/time -v cargo doc --no-deps -p termwiz
...
        Command being timed: "cargo doc --no-deps -p termwiz"
        User time (seconds): 90.74
        System time (seconds): 6.77
        Percent of CPU this job got: 456%
        Elapsed (wall clock) time (h:mm:ss or m:ss): 0:21.37
        Average shared text size (kbytes): 0
        Average unshared data size (kbytes): 0
        Average stack size (kbytes): 0
        Average total size (kbytes): 0
        Maximum resident set size (kbytes): 3810736
        Average resident set size (kbytes): 0
        Major (requiring I/O) page faults: 0
        Minor (reclaiming a frame) page faults: 2504493
        Voluntary context switches: 4301
        Involuntary context switches: 50046
        Swaps: 0
        File system inputs: 0
        File system outputs: 705048
        Socket messages sent: 0
        Socket messages received: 0
        Signals delivered: 0
        Page size (bytes): 4096
        Exit status: 0

Originally posted by @wez in #842 (comment)

[jyn514]

This needs changes to rustwide, see #842 (comment)

[pietroalbini]

If the goal is just to alert the user whether the build failed due to a OOM or a timeout, we don't need to wrap the build with time: Rustwide already returns those reasons when a build fails.

We could change the build_status column in the database from a boolean to an enum/string:

• succeeded
• failed
• failed-oom
• failed-timeout

...and then provide a UI in the build log page displaying a warning if the status is failed-oom or failed-timeout`.

[Nemo157]

Even for successful builds, it'd be nice to record the duration and max-memory usage and show those on the build page.

[wambu-i]

Can I work on this?

[jyn514]

@wambu-i go for it! I think the steps will looks something like this:

1. command.run() gives you a CommandError when the build fails:
   

   docs.rs/src/docbuilder/rustwide_builder.rs

   Lines 527 to 531 in 38ba774

   	let successful = logging::capture(&storage, || {
   	self.prepare_command(build, target, metadata, limits, rustdoc_flags)
   	.and_then(|command| command.run().map_err(failure::Error::from))
   	.is_ok()
   	});

2. Add space for it in BuildResult - I'd change successful: bool to build_error: Option<CommandError> and change all the current successful uses to build_error.is_none().
3. Add a new migration to the database which stores the reason it failed:

   docs.rs/src/db/migrate.rs

   Line 442 in 38ba774

   migration!(

4. Update the builds page to have the new info:

   docs.rs/templates/crate/builds.html

   Lines 44 to 50 in 38ba774

   	<div class="pure-u-1 pure-u-sm-1-24 build">
   	{%- if build.build_status -%}
   	{{ "check" | fas }}
   	{%- else -%}
   	{{ "times" | fas }}
   	{%- endif -%}
   	</div>

5. That template is rendered at

   docs.rs/src/web/crate_details.rs

   Line 261 in 38ba774

   fn map_to_release(conn: &mut Client, crate_id: i32, version: semver::Version) -> Release {

Recording this info for successful builds will need rustwide changes, so let's leave those for later.

#1072 made similar changes and might be helpful.