view tests/test-lfs-serve.t @ 37518:092eff6833a7

lfs: infer the blob store URL from paths.default

If `lfs.url` is specified, it takes precedence.  However, now that we support
serving blobs via hgweb, we shouldn't *require* this setting.  Less
configuration is better (things will work out of the box once this is sorted
out), and git has similar functionality.

This is not a complete solution- it isn't able to infer the blob store from an
explicitly supplied path, and it should consider `paths.default-push` for push.
The pull solution for that is a bit hacky, and this alone is an improvement for
the vast majority of cases.

Even though there are only a handful of references to the saved remote store,
the location of them makes things complicated.

  1) downloading files on demand in the revlog flag processor
  2) copying to readonlyvfs with bundlerepo
  3) downloading in the file prefetch hook
  4) the canupload()/skipdownload() checks
  5) uploading blobs

Since revlog doesn't have a repo or ui reference, we can't avoid creating a
remote store when the extension is loaded.  While the long term goal is to make
sure the prefetch hook is invoked early for every command for efficiency, this
handling in the flag processor is needed as a last ditch fetch.

In order to support the clone command, the remote store needs to be created
later than when the extension loads, since `paths.default` isn't set until just
before the files are checked out.  Therefore, this patch changes the prefetch
hook to ignore the saved reference, and build a new one.

The canupload()/skipdownload() checks simply check if the stored instance is a
`_nullremote`.  Since this can only be set via `lfs.url` (which is reflected in
the saved reference), checking only the instance created when the extension
loaded is fine.

The blob uploading function is called from several places:

  1) a prepush hook
  2) when writing a new bundle
  3) from infinitepush

The prepush hook gets an exchange.pushop, so it has a path to where the push is
going.  The bundle writer and infinitepush don't.  Further, bundle creation for
things like strip and amend are causing blobs to be uploaded.  This seems wrong,
but I don't want to side track this sorting that out, so punt on trying to
handle explicit push paths or `paths.default-push`.

I also think that sending blobs to a remote store when pushing to a local repo
is wrong.  This functionality predates the usercache, so perhaps that's the
reason for it.  I've got some patches floating around to stop sending blobs
remotely in this case, and instead write directly to the other repo's blob
store.  But the tests for corruption handling weren't happy with this change,
and I don't have time to rewrite them.  So exclude filesystem based paths from
this for now.

I don't think there's much of a chance to implement `paths.remote:lfsurl` style
configs, given how early these are resolved vs how late the remote store is
created.  But git has it, so I threw a TODO in there, in case anyone has ideas.

I have no idea why this is now doing http auth twice when it wasn't before.  I
don't think the original blobstore's url is ever being used in these cases.

#testcases lfsremote-on lfsremote-off
#require serve no-reposimplestore

This test splits `hg serve` with and without using the extension into separate
tests cases.  The tests are broken down as follows, where "LFS"/"No-LFS"
indicates whether or not there are commits that use an LFS file, and "D"/"E"
indicates whether or not the extension is loaded.  The "X" cases are not tested
individually, because the lfs requirement causes the process to bail early if
the extension is disabled.

.                        Server
.
.                    No-LFS        LFS
.            +----------------------------+
.            |   ||  D  |  E  |  D  |  E  |
.            |---++=======================|
.  C         | D || N/A | #1  |  X  | #4  |
.  l    No   +---++-----------------------|
.  i    LFS  | E || #2  | #2  |  X  | #5  |
.  e         +---++-----------------------|
.  n         | D ||  X  |  X  |  X  |  X  |
.  t    LFS  |---++-----------------------|
.            | E || #3  | #3  |  X  | #6  |
.            |---++-----------------------+

  $ hg init server
  $ SERVER_REQUIRES="$TESTTMP/server/.hg/requires"

Skip the experimental.changegroup3=True config.  Failure to agree on this comes
first, and causes a "ValueError: no common changegroup version" or "abort:
HTTP Error 500: Internal Server Error", if the extension is only loaded on one
side.  If that *is* enabled, the subsequent failure is "abort: missing processor
for flag '0x2000'!" if the extension is only loaded on one side (possibly also
masked by the Internal Server Error message).
  $ cat >> $HGRCPATH <<EOF
  > [lfs]
  > usercache = null://
  > threshold=10
  > [web]
  > allow_push=*
  > push_ssl=False
  > EOF

#if lfsremote-on
  $ hg --config extensions.lfs= -R server \
  >    serve -p $HGPORT -d --pid-file=hg.pid --errorlog=$TESTTMP/errors.log
#else
  $ hg --config extensions.lfs=! -R server \
  >    serve -p $HGPORT -d --pid-file=hg.pid --errorlog=$TESTTMP/errors.log
#endif

  $ cat hg.pid >> $DAEMON_PIDS
  $ hg clone -q http://localhost:$HGPORT client
  $ grep 'lfs' client/.hg/requires $SERVER_REQUIRES
  [1]

--------------------------------------------------------------------------------
Case #1: client with non-lfs content and the extension disabled; server with
non-lfs content, and the extension enabled.

  $ cd client
  $ echo 'non-lfs' > nonlfs.txt
  $ hg ci -Aqm 'non-lfs'
  $ grep 'lfs' .hg/requires $SERVER_REQUIRES
  [1]

#if lfsremote-on

  $ hg push -q
  $ grep 'lfs' .hg/requires $SERVER_REQUIRES
  [1]

  $ hg clone -q http://localhost:$HGPORT $TESTTMP/client1_clone
  $ grep 'lfs' $TESTTMP/client1_clone/.hg/requires $SERVER_REQUIRES
  [1]

  $ hg init $TESTTMP/client1_pull
  $ hg -R $TESTTMP/client1_pull pull -q http://localhost:$HGPORT
  $ grep 'lfs' $TESTTMP/client1_pull/.hg/requires $SERVER_REQUIRES
  [1]

  $ hg identify http://localhost:$HGPORT
  d437e1d24fbd

#endif

--------------------------------------------------------------------------------
Case #2: client with non-lfs content and the extension enabled; server with
non-lfs content, and the extension state controlled by #testcases.

  $ cat >> $HGRCPATH <<EOF
  > [extensions]
  > lfs =
  > EOF
  $ echo 'non-lfs' > nonlfs2.txt
  $ hg ci -Aqm 'non-lfs file with lfs client'

Since no lfs content has been added yet, the push is allowed, even when the
extension is not enabled remotely.

  $ hg push -q
  $ grep 'lfs' .hg/requires $SERVER_REQUIRES
  [1]

  $ hg clone -q http://localhost:$HGPORT $TESTTMP/client2_clone
  $ grep 'lfs' $TESTTMP/client2_clone/.hg/requires $SERVER_REQUIRES
  [1]

  $ hg init $TESTTMP/client2_pull
  $ hg -R $TESTTMP/client2_pull pull -q http://localhost:$HGPORT
  $ grep 'lfs' $TESTTMP/client2_pull/.hg/requires $SERVER_REQUIRES
  [1]

  $ hg identify http://localhost:$HGPORT
  1477875038c6

--------------------------------------------------------------------------------
Case #3: client with lfs content and the extension enabled; server with
non-lfs content, and the extension state controlled by #testcases.  The server
should have an 'lfs' requirement after it picks up its first commit with a blob.

  $ echo 'this is a big lfs file' > lfs.bin
  $ hg ci -Aqm 'lfs'
  $ grep 'lfs' .hg/requires $SERVER_REQUIRES
  .hg/requires:lfs

#if lfsremote-off
  $ hg push -q
  abort: required features are not supported in the destination: lfs
  (enable the lfs extension on the server)
  [255]
#else
  $ hg push -q
#endif
  $ grep 'lfs' .hg/requires $SERVER_REQUIRES
  .hg/requires:lfs
  $TESTTMP/server/.hg/requires:lfs (lfsremote-on !)

  $ hg clone -q http://localhost:$HGPORT $TESTTMP/client3_clone
  $ grep 'lfs' $TESTTMP/client3_clone/.hg/requires $SERVER_REQUIRES || true
  $TESTTMP/client3_clone/.hg/requires:lfs (lfsremote-on !)
  $TESTTMP/server/.hg/requires:lfs (lfsremote-on !)

  $ hg init $TESTTMP/client3_pull
  $ hg -R $TESTTMP/client3_pull pull -q http://localhost:$HGPORT
  $ grep 'lfs' $TESTTMP/client3_pull/.hg/requires $SERVER_REQUIRES || true
  $TESTTMP/client3_pull/.hg/requires:lfs (lfsremote-on !)
  $TESTTMP/server/.hg/requires:lfs (lfsremote-on !)

The difference here is the push failed above when the extension isn't
enabled on the server.
  $ hg identify http://localhost:$HGPORT
  8374dc4052cb (lfsremote-on !)
  1477875038c6 (lfsremote-off !)

Don't bother testing the lfsremote-off cases- the server won't be able
to launch if there's lfs content and the extension is disabled.

#if lfsremote-on

--------------------------------------------------------------------------------
Case #4: client with non-lfs content and the extension disabled; server with
lfs content, and the extension enabled.

  $ cat >> $HGRCPATH <<EOF
  > [extensions]
  > lfs = !
  > EOF

  $ hg init $TESTTMP/client4
  $ cd $TESTTMP/client4
  $ cat >> .hg/hgrc <<EOF
  > [paths]
  > default = http://localhost:$HGPORT
  > EOF
  $ echo 'non-lfs' > nonlfs2.txt
  $ hg ci -Aqm 'non-lfs'
  $ grep 'lfs' .hg/requires $SERVER_REQUIRES
  $TESTTMP/server/.hg/requires:lfs

  $ hg push -q --force
  warning: repository is unrelated
  $ grep 'lfs' .hg/requires $SERVER_REQUIRES
  $TESTTMP/server/.hg/requires:lfs

TODO: fail more gracefully.

  $ hg clone -q http://localhost:$HGPORT $TESTTMP/client4_clone
  abort: HTTP Error 500: Internal Server Error
  [255]
  $ grep 'lfs' $TESTTMP/client4_clone/.hg/requires $SERVER_REQUIRES
  grep: $TESTTMP/client4_clone/.hg/requires: $ENOENT$
  $TESTTMP/server/.hg/requires:lfs
  [2]

TODO: fail more gracefully.

  $ hg init $TESTTMP/client4_pull
  $ hg -R $TESTTMP/client4_pull pull -q http://localhost:$HGPORT
  abort: HTTP Error 500: Internal Server Error
  [255]
  $ grep 'lfs' $TESTTMP/client4_pull/.hg/requires $SERVER_REQUIRES
  $TESTTMP/server/.hg/requires:lfs

  $ hg identify http://localhost:$HGPORT
  03b080fa9d93

--------------------------------------------------------------------------------
Case #5: client with non-lfs content and the extension enabled; server with
lfs content, and the extension enabled.

  $ cat >> $HGRCPATH <<EOF
  > [extensions]
  > lfs =
  > EOF
  $ echo 'non-lfs' > nonlfs3.txt
  $ hg ci -Aqm 'non-lfs file with lfs client'

  $ hg push -q
  $ grep 'lfs' .hg/requires $SERVER_REQUIRES
  $TESTTMP/server/.hg/requires:lfs

  $ hg clone -q http://localhost:$HGPORT $TESTTMP/client5_clone
  $ grep 'lfs' $TESTTMP/client5_clone/.hg/requires $SERVER_REQUIRES
  $TESTTMP/client5_clone/.hg/requires:lfs
  $TESTTMP/server/.hg/requires:lfs

  $ hg init $TESTTMP/client5_pull
  $ hg -R $TESTTMP/client5_pull pull -q http://localhost:$HGPORT
  $ grep 'lfs' $TESTTMP/client5_pull/.hg/requires $SERVER_REQUIRES
  $TESTTMP/client5_pull/.hg/requires:lfs
  $TESTTMP/server/.hg/requires:lfs

  $ hg identify http://localhost:$HGPORT
  c729025cc5e3

--------------------------------------------------------------------------------
Case #6: client with lfs content and the extension enabled; server with
lfs content, and the extension enabled.

  $ echo 'this is another lfs file' > lfs2.txt
  $ hg ci -Aqm 'lfs file with lfs client'

  $ hg push -q
  $ grep 'lfs' .hg/requires $SERVER_REQUIRES
  .hg/requires:lfs
  $TESTTMP/server/.hg/requires:lfs

  $ hg clone -q http://localhost:$HGPORT $TESTTMP/client6_clone
  $ grep 'lfs' $TESTTMP/client6_clone/.hg/requires $SERVER_REQUIRES
  $TESTTMP/client6_clone/.hg/requires:lfs
  $TESTTMP/server/.hg/requires:lfs

  $ hg init $TESTTMP/client6_pull
  $ hg -R $TESTTMP/client6_pull pull -q http://localhost:$HGPORT
  $ grep 'lfs' $TESTTMP/client6_pull/.hg/requires $SERVER_REQUIRES
  $TESTTMP/client6_pull/.hg/requires:lfs
  $TESTTMP/server/.hg/requires:lfs

  $ hg identify http://localhost:$HGPORT
  d3b84d50eacb

--------------------------------------------------------------------------------
Misc: process dies early if a requirement exists and the extension is disabled

  $ hg --config extensions.lfs=! summary
  abort: repository requires features unknown to this Mercurial: lfs!
  (see https://mercurial-scm.org/wiki/MissingRequirement for more information)
  [255]

#endif

  $ $PYTHON $TESTDIR/killdaemons.py $DAEMON_PIDS

#if lfsremote-on
  $ cat $TESTTMP/errors.log | grep '^[A-Z]'
  Traceback (most recent call last):
  ValueError: no common changegroup version
  Traceback (most recent call last):
  ValueError: no common changegroup version
#else
  $ cat $TESTTMP/errors.log
#endif