[llvm-dev] [cfe-dev] Phabricator -> GitHub PRs?

[David Chisnall via llvm-dev]

On 08/01/2020 00:33, Finkel, Hal J. via cfe-dev wrote:
> As you might imagine, not everyone agrees with you. My thoughts on how 
> to move forward were here: 
> http://lists.llvm.org/pipermail/llvm-dev/2019-November/136591.html - and 
> I do think that we should move forward. It will take some work, however.

I would disagree with your characterisation that Phabricator is a 
superior tool:

Phabricator implements a large superset of the features that most users 
want.  This extra complexity provides a barrier to entry for a lot of 
users (in fact, this morning I had a colleague drop into my office for 
help having done something wrong with Phabricator and mangled the diff, 
and I was unable to help him).

GitHub PRs implement a subset of the functionality that some people 
want.  This causes a problem for people wanting who rely on the other 
features, most commonly reviewers.

Phabricator has the same problem as git: the learning curve is steep and 
there are always new things to learn.  It also has the same advantage as 
git: it probably can do anything that you want it to do, as long as 
you're willing to invest the time to learn.

Favouring Phabricator over GitHub PRs is a decision to prioritise ease 
of use for some reviewers' workflows over that of patch contributors. 
That's fine, if it's a conscious decision (and not even one that I'd 
necessarily disagree with: code reviewers are probably the most scarce 
resource in the LLVM ecosystem and I'd be willing to lose some potential 
contributors if it increased the likelihood of timely and thorough code 
review), but it is misleading to claim that one tool is 'superior' in 
the abstract, without defining the requirements.

David