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

David Greene via llvm-dev llvm-dev at lists.llvm.org
Thu Jan 16 10:03:04 PST 2020

"Doerfert, Johannes" <jdoerfert at anl.gov> writes:

>> The above journey to get a patch submitted was my actual first
>> experience with Phab.  It did not leave a good taste in my mouth.
> What if we take these points and act on them?

That would be great!

>  It should be "fairly simple" to simplify the interface by hiding less
>  often used apps.

This brings another question to mind.  How much does it cost us to host
our own Phab instance and customize/maintain it?  Would we save anything
by going to a third-party hosted solution like we did for the
repository?  Note that I am not saying that third-party hosted review
tool must be GitHub.  It could be something else, though GitHub is an
obvious choice since the repository already lives there.

>  While writing documentation is not easy, it would also help I guess.

Documentation would help.  In particular I would like some guidance on
when a patch series is appropriate vs. individual patches submitted for
review.  I haven't been able to find any documentation on how that model
works in Phab, when you would want to use it, the features it offers,
etc.  It may exist on phabricator.com but it's not obvious where it is.

It is clear to me that I am missing a lot of potential usefulness with
Phab and I would like to learn more!

One thing I *really* miss is a command-line tool to submit patches.
With GitHub you just use git (there are even extension commands for
opening GitHub pull requests).  Easy-peasy.

Mozilla's Phab tools may fill that need.  I am optimistic here.  If they
work well the project should consider making them an officially
supported part of our tool belt with LLVM-specific documentation on how
to use them.

I am certainly open to sticking with Phab if we can address its issues.
It will be important to solicit input from those very new to the project
as many of us have figured out how to work around its quirks to the
point that we don't even recognize the issues anymore.


More information about the llvm-dev mailing list