changegroup: remove reordering control (BC)
This logic - including the experimental bundle.reorder option -
was originally added in
a8e3931e3fb5 in 2011 and then later ported
to changegroup.py.
The intent of this option and associated logic is to control
the ordering of revisions in deltagroups in changegroups. At the
time it was implemented, only changegroup version 1 existed
and generaldelta revlogs were just coming into the world. Changegroup
version 1 requires that deltas be made against the last revision
sent over the wire. Used with generaldelta, this created an
impedance mismatch of sorts and resulted in changegroup producers
spending a lot of time recomputing deltas.
Revision reordering was introduced so outgoing revisions would be
sent in "generaldelta order" and producers would be able to
reuse internal deltas from storage.
Later on, we introduced changegroup version 2. It supported denoting
which revision a delta was against. So we no longer needed to
sort outgoing revisions to ensure optimal delta generation from the
producer. So, subsequent changegroup versions disabled reordering.
We also later made the changelog not store deltas by default. And
we also made the changelog send out deltas in storage order. Why we
do this for changelog, I'm not sure. Maybe we want to preserve revision
order across clones? It doesn't really matter for this commit.
Fast forward to 2018. We want to abstract storage backends. And having
changegroup code require knowledge about how deltas are stored
internally interferes with that goal.
This commit removes reordering control from changegroup generation.
After this commit, the reordering behavior is:
* The changelog is always sent out in storage order (no behavior
change).
* Non-changelog generaldelta revlogs are reordered to always be in DAG
topological order (previously, generaldelta revlogs would be emitted
in storage order for version 2 and 3 changegroups).
* Non-changelog non-generaldelta revlogs are sent in storage order (no
behavior change).
* There exists no config option to override behavior.
The big difference here is that generaldelta revlogs now *always* have
their revisions sorted in DAG order before going out over the wire. This
behavior was previously only done for changegroup version 1. Version 2
and version 3 changegroups disabled reordering because the interchange
format supported encoding arbitrary delta parents, so reordering wasn't
strictly necessary.
I can think of a few significant implications for this change.
Because changegroup receivers will now see non-changelog revisions
in DAG order instead of storage order, the internal storage order of
manifests and files may differ substantially between producer and
consumer. I don't think this matters that much, since the storage
order of manifests and files is largely hidden from users. Only
the storage order of changelog matters (because `hg log` shows the
changelog in storage order). I don't think there should be any
controversy here.
The reordering of revisions has implications for changegroup producers.
Previously, generaldelta revlogs would be emitted in storage order.
And in the common case, the internally-stored delta could effectively
be copied from disk into the deltagroup delta. This meant that emitting
delta groups for generaldelta revlogs would be mostly linear read I/O.
This is desirable for performance. With us now reordering generaldelta
revlog revisions in DAG order, the read operations may use more random
I/O instead of sequential I/O. This could result in performance
loss. But with the prevalence of SSDs and fast random I/O, I'm not
too worried. (Note: the optimal emission order for revlogs is actually
delta encoding order. But the changegroup code wasn't doing that before
or after this change. We could potentially implement that in a later
commit.)
Changegroups in DAG order will have implications for receivers.
Previously, receiving storage order might mean seeing a number of
interleaved branches. This would mean long delta chains, sparse
I/O, and possibly more fulltext revisions instead of deltas, blowing
up storage storage. (This is the same set of problems that sparse
revlogs aims to address.) With the producer now sending revisions in DAG
order, the receiver also stores revisions in DAG order. That means
revisions for the same DAG branch are all grouped together. And this
should yield better storage outcomes. In other words, sending the
reordered changegroup allows the receiver to have better storage
order and for the producer to not propagate its (possibly sub-optimal)
internal storage order.
On the mozilla-unified repository, this change influences bundle
generation:
$ hg bundle -t none-v2 -a
before: time: real 355.680 secs (user 256.790+0.000 sys 16.820+0.000)
after: time: real 382.950 secs (user 281.700+0.000 sys 17.690+0.000)
before: 7,150,228,967 bytes (uncompressed)
after: 7,041,556,273 bytes (uncompressed)
before: 1,669,063,234 bytes (zstd l=3)
after: 1,628,598,830 bytes (zstd l=3)
$ hg unbundle
before: time: real 511.910 secs (user 466.750+0.000 sys 32.680+0.000)
after: time: real 487.790 secs (user 443.940+0.000 sys 30.840+0.000)
00manifest.d size:
source: 274,924,292 bytes
before: 304,741,626 bytes
after: 245,252,087 bytes
.hg/store total file size:
source: 2,649,133,490
before: 2,680,888,130
after: 2,627,875,673
We see the bundle size drop. That's probably because if a revlog
internally isn't storing a delta, it will choose to delta against
the last emitted revision. And on repos with interleaved branches
(like mozilla-unified), the previous revision could be an unrelated
branch and therefore be a large delta. But with this patch, the
previous revision is likely p1 or p2 and a delta should be small.
We also see the manifest size drop by ~50 MB. It's worth noting that
the manifest actually *increased* in size by ~25 MB in the old
strategy and decreased ~25 MB from its source in the new strategy.
Again, my explanation for this is that the DAG ordering in the
changegroup is resulting in better grouping of revisions in the
receiver, which results in more compact delta chains and higher
storage efficiency.
Unbundle time also dropped. I suspect this is due to the revlog having
to work less to compute deltas since the incoming deltas are more
optimal. i.e. the receiver spends less time resolving fulltext
revisions as incoming deltas bounce around between DAG branches and
delta chains.
We also see bundle generation time increase. This is not desirable.
However, the regression is only significant on the original repository:
if we generate a bundle from the repository created from the new,
always reordered bundles, we're close to baseline (if not at it with
expected noise):
$ hg bundle -t none-v2 -a
before (original): time: real 355.680 secs (user 256.790+0.000 sys 16.820+0.000)
after (original): time: real 382.950 secs (user 281.700+0.000 sys 17.690+0.000)
after (new repo): time: real 362.280 secs (user 260.300+0.000 sys 17.700+0.000)
This regression is a bit worrying because it will impact serving
canonical repositories (that don't have optimal internal storage
unless they are reordered - possibly as part of running
`hg debugupgraderepo`). However, this regression will only be
noticed by very large changegroups. And I'm guessing/hoping that
any repository that large is using clonebundles to mitigate server
load.
Again, sending DAG order isn't the optimal send order for servers:
sending in storage-delta order is. But in order to enable
storage-optimal send order, we'll need a storage API that handles
sorting. Future commits will introduce such an API.
Differential Revision: https://phab.mercurial-scm.org/D4721
$ echo "[extensions]" >> $HGRCPATH
$ echo "mq=" >> $HGRCPATH
$ hg init
$ hg qinit
$ echo x > x
$ hg ci -Ama
adding x
$ hg qnew a.patch
$ echo a > a
$ hg add a
$ hg qrefresh
$ hg qnew b.patch
$ echo b > b
$ hg add b
$ hg qrefresh
$ hg qnew c.patch
$ echo c > c
$ hg add c
$ hg qrefresh
$ hg qpop -a
popping c.patch
popping b.patch
popping a.patch
patch queue now empty
should fail
$ hg qguard does-not-exist.patch +bleh
abort: no patch named does-not-exist.patch
[255]
should fail
$ hg qguard +fail
abort: no patches applied
[255]
$ hg qpush
applying a.patch
now at: a.patch
should guard a.patch
$ hg qguard +a
should print +a
$ hg qguard
a.patch: +a
$ hg qpop
popping a.patch
patch queue now empty
should fail
$ hg qpush a.patch
cannot push 'a.patch' - guarded by '+a'
[1]
$ hg qguard a.patch
a.patch: +a
should push b.patch
$ hg qpush
applying b.patch
now at: b.patch
$ hg qpop
popping b.patch
patch queue now empty
test selection of an empty guard
$ hg qselect ""
abort: guard cannot be an empty string
[255]
$ hg qselect a
number of unguarded, unapplied patches has changed from 2 to 3
should push a.patch
$ hg qpush
applying a.patch
now at: a.patch
$ hg qguard -- c.patch -a
should print -a
$ hg qguard c.patch
c.patch: -a
should skip c.patch
$ hg qpush -a
applying b.patch
skipping c.patch - guarded by '-a'
now at: b.patch
$ hg qnext
all patches applied
[1]
should display b.patch
$ hg qtop
b.patch
$ hg qguard -n c.patch
should push c.patch
$ hg qpush -a
applying c.patch
now at: c.patch
$ hg qpop -a
popping c.patch
popping b.patch
popping a.patch
patch queue now empty
$ hg qselect -n
guards deactivated
number of unguarded, unapplied patches has changed from 3 to 2
should push all
$ hg qpush -a
applying b.patch
applying c.patch
now at: c.patch
$ hg qpop -a
popping c.patch
popping b.patch
patch queue now empty
$ hg qguard a.patch +1
$ hg qguard b.patch +2
$ hg qselect 1
number of unguarded, unapplied patches has changed from 1 to 2
should push a.patch, not b.patch
$ hg qpush
applying a.patch
now at: a.patch
$ hg qpush
applying c.patch
now at: c.patch
$ hg qpop -a
popping c.patch
popping a.patch
patch queue now empty
$ hg qselect 2
should push b.patch
$ hg qpush
applying b.patch
now at: b.patch
$ hg qpush -a
applying c.patch
now at: c.patch
$ hg qprev
b.patch
Used to be an issue with holes in the patch sequence
So, put one hole on the base and ask for topmost patch.
$ hg qtop
c.patch
$ hg qpop -a
popping c.patch
popping b.patch
patch queue now empty
$ hg qselect 1 2
number of unguarded, unapplied patches has changed from 2 to 3
should push a.patch, b.patch
$ hg qpush
applying a.patch
now at: a.patch
$ hg qpush
applying b.patch
now at: b.patch
$ hg qpop -a
popping b.patch
popping a.patch
patch queue now empty
$ hg qguard -- a.patch +1 +2 -3
$ hg qselect 1 2 3
number of unguarded, unapplied patches has changed from 3 to 2
list patches and guards
$ hg qguard -l
a.patch: +1 +2 -3
b.patch: +2
c.patch: unguarded
have at least one patch applied to test coloring
$ hg qpush
applying b.patch
now at: b.patch
list patches and guards with color
$ hg --config extensions.color= qguard --config color.mode=ansi \
> -l --color=always
\x1b[0;30;1ma.patch\x1b[0m: \x1b[0;33m+1\x1b[0m \x1b[0;33m+2\x1b[0m \x1b[0;31m-3\x1b[0m (esc)
\x1b[0;34;1;4mb.patch\x1b[0m: \x1b[0;33m+2\x1b[0m (esc)
\x1b[0;30;1mc.patch\x1b[0m: \x1b[0;32munguarded\x1b[0m (esc)
should pop b.patch
$ hg qpop
popping b.patch
patch queue now empty
list series
$ hg qseries -v
0 G a.patch
1 U b.patch
2 U c.patch
list guards
$ hg qselect
1
2
3
should push b.patch
$ hg qpush
applying b.patch
now at: b.patch
$ hg qpush -a
applying c.patch
now at: c.patch
$ hg qselect -n --reapply -v
guards deactivated
popping guarded patches
popping c.patch
popping b.patch
patch queue now empty
reapplying unguarded patches
skipping a.patch - guarded by '+1' '+2'
skipping b.patch - guarded by '+2'
skipping a.patch - guarded by '+1' '+2'
skipping b.patch - guarded by '+2'
applying c.patch
patching file c
adding c
committing files:
c
committing manifest
committing changelog
now at: c.patch
guards in series file: +1 +2 -3
$ hg qselect -s
+1
+2
-3
should show c.patch
$ hg qapplied
c.patch
$ hg qrename a.patch new.patch
should show :
new.patch: +1 +2 -3
b.patch: +2
c.patch: unguarded
$ hg qguard -l
new.patch: +1 +2 -3
b.patch: +2
c.patch: unguarded
$ hg qnew d.patch
$ hg qpop
popping d.patch
now at: c.patch
should show new.patch and b.patch as Guarded, c.patch as Applied
and d.patch as Unapplied
$ hg qseries -v
0 G new.patch
1 G b.patch
2 A c.patch
3 U d.patch
qseries again, but with color
$ hg --config extensions.color= --config color.mode=ansi qseries -v --color=always
0 G \x1b[0;30;1mnew.patch\x1b[0m (esc)
1 G \x1b[0;30;1mb.patch\x1b[0m (esc)
2 A \x1b[0;34;1;4mc.patch\x1b[0m (esc)
3 U \x1b[0;30;1md.patch\x1b[0m (esc)
$ hg qguard d.patch +2
new.patch, b.patch: Guarded. c.patch: Applied. d.patch: Guarded.
$ hg qseries -v
0 G new.patch
1 G b.patch
2 A c.patch
3 G d.patch
$ qappunappv()
> {
> for command in qapplied "qapplied -v" qunapplied "qunapplied -v"; do
> echo % hg $command
> hg $command
> done
> }
$ hg qpop -a
popping c.patch
patch queue now empty
$ hg qguard -l
new.patch: +1 +2 -3
b.patch: +2
c.patch: unguarded
d.patch: +2
$ qappunappv
% hg qapplied
% hg qapplied -v
% hg qunapplied
c.patch
% hg qunapplied -v
0 G new.patch
1 G b.patch
2 U c.patch
3 G d.patch
$ hg qselect 1
number of unguarded, unapplied patches has changed from 1 to 2
$ qappunappv
% hg qapplied
% hg qapplied -v
% hg qunapplied
new.patch
c.patch
% hg qunapplied -v
0 U new.patch
1 G b.patch
2 U c.patch
3 G d.patch
$ hg qpush -a
applying new.patch
skipping b.patch - guarded by '+2'
applying c.patch
skipping d.patch - guarded by '+2'
now at: c.patch
$ qappunappv
% hg qapplied
new.patch
c.patch
% hg qapplied -v
0 A new.patch
1 G b.patch
2 A c.patch
% hg qunapplied
% hg qunapplied -v
3 G d.patch
$ hg qselect 2
number of unguarded, unapplied patches has changed from 0 to 1
$ qappunappv
% hg qapplied
new.patch
c.patch
% hg qapplied -v
0 A new.patch
1 U b.patch
2 A c.patch
% hg qunapplied
d.patch
% hg qunapplied -v
3 U d.patch
$ for patch in `hg qseries`; do
> echo % hg qapplied $patch
> hg qapplied $patch
> echo % hg qunapplied $patch
> hg qunapplied $patch
> done
% hg qapplied new.patch
new.patch
% hg qunapplied new.patch
b.patch
d.patch
% hg qapplied b.patch
new.patch
% hg qunapplied b.patch
d.patch
% hg qapplied c.patch
new.patch
c.patch
% hg qunapplied c.patch
d.patch
% hg qapplied d.patch
new.patch
c.patch
% hg qunapplied d.patch
hg qseries -m: only b.patch should be shown
the guards file was not ignored in the past
$ hg qdelete -k b.patch
$ hg qseries -m
b.patch
hg qseries -m with color
$ hg --config extensions.color= --config color.mode=ansi qseries -m --color=always
\x1b[0;31;1mb.patch\x1b[0m (esc)
excercise corner cases in "qselect --reapply"
$ hg qpop -a
popping c.patch
popping new.patch
patch queue now empty
$ hg qguard -- new.patch -not-new
$ hg qguard -- c.patch -not-c
$ hg qguard -- d.patch -not-d
$ hg qpush -a
applying new.patch
applying c.patch
applying d.patch
patch d.patch is empty
now at: d.patch
$ hg qguard -l
new.patch: -not-new
c.patch: -not-c
d.patch: -not-d
$ hg qselect --reapply not-d
popping guarded patches
popping d.patch
now at: c.patch
reapplying unguarded patches
cannot push 'd.patch' - guarded by '-not-d'
$ hg qser -v
0 A new.patch
1 A c.patch
2 G d.patch
$ hg qselect --reapply -n
guards deactivated
$ hg qpush
applying d.patch
patch d.patch is empty
now at: d.patch
$ hg qser -v
0 A new.patch
1 A c.patch
2 A d.patch
$ hg qselect --reapply not-c
popping guarded patches
popping d.patch
popping c.patch
now at: new.patch
reapplying unguarded patches
applying d.patch
patch d.patch is empty
now at: d.patch
$ hg qser -v
0 A new.patch
1 G c.patch
2 A d.patch
$ hg qselect --reapply not-new
popping guarded patches
popping d.patch
popping new.patch
patch queue now empty
reapplying unguarded patches
applying c.patch
applying d.patch
patch d.patch is empty
now at: d.patch
$ hg qser -v
0 G new.patch
1 A c.patch
2 A d.patch
test that qselect shows "number of guarded, applied patches" correctly
$ hg qimport -q -e b.patch
adding b.patch to series file
$ hg qguard -- b.patch -not-b
$ hg qpop -a -q
patch queue now empty
$ hg qunapplied -v
0 G new.patch
1 U c.patch
2 U d.patch
3 U b.patch
$ hg qselect not-new not-c
number of unguarded, unapplied patches has changed from 3 to 2
$ hg qpush -q -a
patch d.patch is empty
now at: b.patch
$ hg qapplied -v
0 G new.patch
1 G c.patch
2 A d.patch
3 A b.patch
$ hg qselect --none
guards deactivated
$ hg qselect not-new not-c not-d
number of guarded, applied patches has changed from 0 to 1
test that "qselect --reapply" reapplies patches successfully when the
already applied patch becomes unguarded and it follows the already
guarded (= not yet applied) one.
$ hg qpop -q -a
patch queue now empty
$ hg qselect not-new not-c
number of unguarded, unapplied patches has changed from 1 to 2
$ hg qpush -q -a
patch d.patch is empty
now at: b.patch
$ hg qapplied -v
0 G new.patch
1 G c.patch
2 A d.patch
3 A b.patch
$ hg qselect -q --reapply not-c not-b
now at: d.patch
cannot push 'b.patch' - guarded by '-not-b'
$ hg qseries -v
0 U new.patch
1 G c.patch
2 A d.patch
3 G b.patch
test that "qselect --reapply" checks applied patches correctly when no
applied patches becomes guarded but some of unapplied ones become
unguarded.
$ hg qpop -q -a
patch queue now empty
$ hg qselect not-new not-c not-d
number of unguarded, unapplied patches has changed from 2 to 1
$ hg qpush -q -a
now at: b.patch
$ hg qapplied -v
0 G new.patch
1 G c.patch
2 G d.patch
3 A b.patch
$ hg qselect -q --reapply not-new not-c
$ hg qseries -v
0 G new.patch
1 G c.patch
2 U d.patch
3 A b.patch