Switch default deference back to True for tarball artifact

Review Request #4304 — Created Oct. 13, 2016 and submitted

benjyw, jsirois, mateor, peiyu, stuhood

https://rbcommons.com/s/twitter/r/4166 set default deference=False, causing bundle failures because bundle cache is a link to a remote location rather than actual the content itself. This RB changes the default back to True while leaving the interface structure for other tasks to consume later on.


  1. Not sure how easy it is to add a test to check bundled artifact from cache are dereferenced

    1. Not hard to add, but this RB is more of a revert. It does make sense to do so when a new change comes in to use the non default values.

      lrwxr-xr-x  1 yic  staff  4 Oct 13 11:00 a -> ../b
      [tw-mbp-yic dummy]$ 
      [tw-mbp-yic dummy]$ cd ..
      [tw-mbp-yic workspace]$ 
      [tw-mbp-yic workspace]$ tar -cvf x.tar dummy
      a dummy
      a dummy/a
      [tw-mbp-yic workspace]$ tar -cvhf y.tar dummy
      a dummy
      a dummy/a
      [tw-mbp-yic workspace]$ tar tvf x.tar 
      drwxr-xr-x  0 yic    staff       0 Oct 13 11:00 dummy/
      lrwxr-xr-x  0 yic    staff       0 Oct 13 11:00 dummy/a -> ../b
      [tw-mbp-yic workspace]$ tar tvf y.tar 
      drwxr-xr-x  0 yic    staff       0 Oct 13 11:00 dummy/
      -rw-r--r--  0 yic    staff       5 Oct 13 10:54 dummy/a
  1. In this case, I wonder whether it would be easier to explicitly copy into the results_dir for bundles?

    What does the failure look like? It would be good to open a ticket with the error you experienced, and include it as a TODO next to dereference=True, as I think that preserving exactly the structure of the cache entry is actually a good idea moving forward.

  1. ...to be clear: fine with the revert; just want to see a ticket. #shipit

Review request changed

Status: Closed (submitted)

Change Summary:

b3b009909df1c513d66e7085f8825c4691e516dd thanks gents!