From fd37a55958705619e045a5ea003f4b95c1caa187 Mon Sep 17 00:00:00 2001 From: Philip Metzger Date: Tue, 2 May 2023 16:30:25 +0200 Subject: [PATCH] WIP: Features which need a implementation Split per feature --- cli/src/commands/run.rs | 56 +++++++++++++++++++++++++++++++++++++++++ 1 file changed, 56 insertions(+) diff --git a/cli/src/commands/run.rs b/cli/src/commands/run.rs index d288e162a7..83b0aa5c36 100644 --- a/cli/src/commands/run.rs +++ b/cli/src/commands/run.rs @@ -166,6 +166,21 @@ async fn rewrite_commit( } } +/// The Strategy `run` uses to deal with any failure in it's subprocesses. +#[derive(Debug, Default, Clone, clap::ValueEnum)] +pub(crate) enum ErrorStrategy { + /// Stop at any revision if the subprocess fails. + /// Cancels any scheduled work but lets already running processes finish + /// their work. + #[default] + Stop, + /// Continue if an error occurs, similar to Make's --keep-going. + Continue, + /// Signal a fatal failure and stop immediately any processes. Also stops + /// any running jobs. + Fatal, +} + /// Run a command across a set of revisions. /// /// @@ -192,6 +207,34 @@ pub struct RunArgs { /// How many processes should run in parallel, uses by default all cores. #[arg(long, short)] jobs: Option, + /// Setup the required environment and instead of applying running script, + /// display all actions which would be done. + #[arg(long, short)] + dry_run: bool, + /// Ignore changes across multiple run invocations. + #[arg(long)] + read_only: bool, + /// Show the diff of an affected revision. + #[arg(long)] + show: Option, + /// After running the command rebase the revisions parent onto the new + /// commit. This is the default strategy. + #[arg(long)] + rebase: bool, + /// After running the command, re-adjust the parent commit to the new + /// revision. + #[arg(long, conflicts_with = "rebase")] + reparent: bool, + /// Clean up all hidden state for run. + #[arg(long)] + clean: bool, + /// Run the command for all revisions, even if a single one fails. + /// Implies error-strategy=continue + #[arg(short, long)] + keep_going: bool, + /// The strategy `jj run` should use, if a failure occurs. + #[arg(value_enum, default_value = "ErrorStrategy::Stop")] + error_strategy: ErrorStrategy, } pub fn cmd_run(ui: &mut Ui, command: &CommandHelper, args: &RunArgs) -> Result<(), CommandError> { @@ -225,6 +268,19 @@ pub fn cmd_run(ui: &mut Ui, command: &CommandHelper, args: &RunArgs) -> Result<( let stored_commits = create_working_copies(topo_sorted_commits)?; let tx = workspace_command.start_transaction(); + let mut workqueue = WorkQueue::builder() + .commits(topo_sorted_commits) + .command(args.shell_command) + // If the user has not specified the amount of jobs to use, create one per cpu core, + // this models falls apart for large jobs, which themselves require a nproc jobs + // like cargo or ninja. + .jobs(_jobs) + .continue_on_error(true) + // .fail_on_error(if opts.fail_fast || other_thing { ... } else {} ) + // .reparent_commits(opts.reparent); + .build(); + let tx = workspace_command.start_transaction(); + // let (error_recv, status_recv) = workqueue.channels(); // Start all the jobs. run_inner(tx, sender_tx, jobs, &args.shell_command, stored_commits) .await?