Restore wd on bundle task failure

[weshinsley]

No description provided.

[weshinsley]

So, the setwd method is not ideal...

What's happening is that in orderly_version.R, around lines 599-605, within the run_execute function:-

 withr::with_envvar(private$envvar, {
        withr::with_dir(private$workdir, {
          source(private$recipe$script, local = private$envir, # nolint
                 echo = echo, max.deparse.length = Inf)
        })
      })

If the source call encounters a stop/crash, then control passes straight to the withCallingHandlers block in the calling function, around line 541-555, without resetting the working directory to what it was before the with_dir ->

  currentwd <- getwd()
      withCallingHandlers({
        ## Refetch the preflight info here: we want to keep git but
        ## replace everything else I think.  We might save the random
        ## seed but that is not actually supposed to work across R
        ## versions.
        private$preflight()
        private$preflight_info["git"] <- info$preflight_info["git"]

        self$run_execute(echo)   # <--------- here's the call within which we crash
        self$run_cleanup()       # <----- so after a crash, skip this line, go straight to the error clause
      }, error = function(e) {
        setwd(currentwd)
        self$run_failed_cleanup(e)
      })

[codecov]

Codecov Report

Merging #325 (569f1d3) into master (d02fc85) will increase coverage by 0.00%.
The diff coverage is 100.00%.

❗ Current head 569f1d3 differs from pull request most recent head 8ee1da0. Consider uploading reports for the commit 8ee1da0 to get more accurate results

@@           Coverage Diff           @@
##           master     #325   +/-   ##
=======================================
  Coverage   99.95%   99.95%           
=======================================
  Files          41       41           
  Lines        4535     4543    +8     
=======================================
+ Hits         4533     4541    +8     
  Misses          2        2           

Impacted Files	Coverage Δ
R/orderly_version.R	100.00% <100.00%> (ø)
R/query_search.R	100.00% <100.00%> (ø)

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[r-ash]

Nice one

[weshinsley]

Thanks Rob - let's merge this and hopefully we can then get more info on Paddy's cluster issue. @richfitz - this ends up being a one-line fix which you did in the end, plus some tests - tagging you here as we mentioned it last week, re whether this is relevant for orderly3, or whether that migt be different enough anyway,.

[richfitz]

thanks guys