Pulkit Goyal <7895pulkit@gmail.com> [Fri, 23 Feb 2018 16:57:17 +0530] rev 36375
py3: add b'' prefixes in test-abort-checkin.t
# skip-blame because we just added a b'' prefix.
Pulkit Goyal <7895pulkit@gmail.com> [Wed, 21 Feb 2018 23:43:23 +0530] rev 36374
py3: add b'' prefixes in test-dispatch.py
# skip-blame because this is just adding b'' prefixes
Augie Fackler <augie@google.com> [Thu, 22 Feb 2018 20:04:42 -0500] rev 36373
cleanup: say goodbye to manifestv2 format
This experiment was a bust: we'd hoped for smaller repository sizes,
but things got larger. Google ended up rolling out tree manifests in a
format that's compatible with the original manifest format, and I
believe Facebook is doing the same. This code was never implemented as
native speedups, so I'm pretty comfortable saying nobody is using the
experimental feature. Let's rip it out.
I noticed this code still kicking around because I was investigating a
repo corruption issue for timeless.
.. bc::
Support for the experimental manifestv2 format has been removed, as
it was never completed and failed to meet expectations.
Differential Revision: https://phab.mercurial-scm.org/D2393
Gregory Szorc <gregory.szorc@gmail.com> [Wed, 21 Feb 2018 16:47:39 -0800] rev 36372
wireproto: document the wonky push protocol for SSH
It took me several minutes to figure out how the "unbundle"
protocol worked. It turns out that the SSH protocol handler
sends an empty reply that is interpreted as "OK to send" and
only then does the client send the bundle payload.
On top of that, the response is different depending on whether
the operation was successful or not. I nearly pulled out my hair
deciphering this.
Differential Revision: https://phab.mercurial-scm.org/D2385