panic when releases.files is NULL

[jyn514]

Caught this from looking through the logs on the build server: https://docs.rs/crate/rips/0.2.0/source/ gives a 500 when you visit it. The build for rips failed, maybe that's why the files are NULL?

Backtrace:

web_1  | thread '<unnamed>' panicked at 'error retrieving column 5: Error(Conversion(WasNull))', /root/.cargo/registry/src/github.com-1ecc6299db9ec823/postgres-0.15.2/src/rows.rs:201:31
web_1  | stack backtrace:
web_1  |   13: postgres::rows::Row::get
web_1  |   14: cratesfyi::web::source::source_browser_handler
web_1  |   15: <F as iron::middleware::Handler>::handle
web_1  |   16: <alloc::boxed::Box<dyn iron::middleware::Handler> as iron::middleware::Handler>::handle
web_1  |   17: router::router::Router::handle_method
web_1  |   18: <router::router::Router as iron::middleware::Handler>::handle
web_1  |   19: iron::middleware::Chain::continue_from_handler
web_1  |   20: <iron::middleware::Chain as iron::middleware::Handler>::handle
web_1  |   21: <alloc::boxed::Box<dyn iron::middleware::Handler> as iron::middleware::Handler>::handle
web_1  |   22: <cratesfyi::web::CratesfyiHandler as iron::middleware::Handler>::handle
web_1  |   23: <iron::iron::RawHandler<H> as hyper::server::Handler>::handle
web_1  |   24: hyper::server::Worker<H>::handle_connection
web_1  |   25: hyper::server::listener::spawn_with::{{closure}}

Note for maintainers: to find other crates with this issue, run this psql command on the server:

SELECT releases.target_name, releases.version FROM releases WHERE releases.files IS NULL;

[jyn514]

It looks like this happens whenever the build fails.

[jyn514]

Relevant code: https://github.com/rust-lang/docs.rs/blob/master/src/web/source.rs#L107

[jyn514]

Hmm ... I worked around this by returning None in the linked code, which avoids the panic, but the rendered page is still mostly blank :(

[jyn514]

I turned it into a 404 instead, at least we're giving an error message now.

[jyn514]

Reopening because of #521

[jyn514]

While fixing this, I noticed that using the current master branch, binaries do not have their source code saved into the database. However, if I go to e.g. https://docs.rs/crate/bat/0.12.1/source/ all the code is there. Maybe the builder got changed at some point? If so I think it makes more sense to store the code in the DB even for binaries than to special case the error handling if it's not there (see also #520).

cc @pietroalbini , could this have happened during the switch to rustwide?

[pietroalbini]

Not sure if I see the point of storing the source code for binaries... people shouldn't look on docs.rs for them.

[jyn514]

Ok, in that case I made a PR to fix the panic.

[jyn514]

Fixed in #526