Release automation aborts without reason

[tlucas]

This is a regression, introduced with #4354 / #4355, git:c6f892d / hg:ef4ad61830eb

How to reproduce

Run the release automation for a setup, which should succeed (no uncommitted / outgoing / incoming changes, see #4353 / #4355) for e.g. adblockpluschrome:

1. Clone the adblockpluschrome repository (or reset your copy to master)
2. Clone the downloads repository (or reset your copy to master)
3. Run the release automation:

./build.py -t chrome release -k <certificate> 1.14.0

(Important: configure fake remotes for your adblockpluschrome and downloads repository)

Observed behaviour

The release aborts.

Expected behaviour

The release should succeed, updating both the adblockpluschrome and downloads repository

Hints for testers

To setup fake remotes for the repositories, you can follow these steps:

• Clone the adblockpluschrome repository to a location of your choice hg clone https://hg.adblockplus.org/adblockpluschrome
• Clone the downloads repository to a location of your choice (i suggest the same location as abp-chrome - otherwise you would have to specify the downloads location in the build process) hg clone https://hg.adblockplus.org/downloads
• Create the fake remotes for adblockpluschrome and downloads:
  • cp -ar adblockpluschrome abpc_remote
  • cp -ar downloads downloads_remote
  • Configure your main repositories to have the fake remotes as remotes: change .hg/hgrc in both main repositories to contain default = <path_to_fake_remote> e.g.: default = /home/tlucas/abp/tmphg/abpc_remote
  • you can verify this by running hg incoming in the main repositories, it should say "comparing with <your_fake_remote>"
• Create your own .pem (this requires openssl to be installed) in a location of your choice: openssl req -newkey rsa:2048 -nodes -x509 -days 365 -out certificate.pem - additionally to the certificate.pem, this creates a key.pem, which is the file that we need.
• run the release in your adblockpluschrome main repository: ./build.py -t chrome release -k <path_to_new_key.pem> 1.14.0

[kzar]

From the discussion in IRC this sounds like a recent regression. Please could you bisect recent changes to figure out which commit is to blame, and then mention that in the description of this issue? That's usually a good place to start with these kind of problems.

[abpbot]

A commit referencing this issue has landed:
​Issue 5736 - Fix causeless abort of release