procutil: make stream detection in make_line_buffered more correct and strict
In make_line_buffered(), we don’t want to wrap the stream if we know that lines
get flushed to the underlying raw stream already.
Previously, the heuristic was too optimistic. It assumed that any stream which
is not an instance of io.BufferedIOBase doesn’t need wrapping. However, there
are buffered streams that aren’t instances of io.BufferedIOBase, like
Mercurial’s own winstdout.
The new logic is different in two ways:
First, only for the check, if unwraps any combination of WriteAllWrapper and
winstdout.
Second, it skips wrapping the stream only if it is an instance of io.RawIOBase
(or already wrapped). If it is an instance of io.BufferedIOBase, it gets
wrapped. In any other case, the function raises an exception. This ensures
that, if an unknown stream is passed or we add another wrapper in the future,
we don’t wrap the stream if it’s already line buffered or not wrap the stream
if it’s not line buffered. In fact, this was already helpful during development
of this change. Without it, I possibly would have forgot that WriteAllWrapper
needs to be ignored for the check, leading to unnecessary wrapping if stdout is
unbuffered.
The alternative would have been to always wrap unknown streams. However, I
don’t think that anyone would benefit from being less strict. We can expect
streams from the standard library to be subclassing either io.RawIOBase or
io.BufferedIOBase, so running Mercurial in the standard way should not regress
by this change. Py2exe might replace sys.stdout and sys.stderr, but that
currently breaks Mercurial anyway and also these streams don’t claim to be
interactive, so this function is not called for them.
Test transaction safety
=======================
#testcases revlogv1 revlogv2 changelogv2
#if revlogv1
$ cat << EOF >> $HGRCPATH
> [experimental]
> revlogv2=no
> EOF
#endif
#if revlogv2
$ cat << EOF >> $HGRCPATH
> [experimental]
> revlogv2=enable-unstable-format-and-corrupt-my-data
> EOF
#endif
#if changelogv2
$ cat << EOF >> $HGRCPATH
> [format]
> exp-use-changelog-v2=enable-unstable-format-and-corrupt-my-data
> EOF
#endif
This test basic case to make sure external process do not see transaction
content until it is committed.
# TODO: also add an external reader accessing revlog files while they are written
# (instead of during transaction finalisation)
# TODO: also add stream clone and hardlink clone happening during these transaction.
setup
-----
synchronisation+output script using the following schedule:
[A1] "external" is started
[A2] "external" waits on EXT_UNLOCK
[A2] "external" + creates EXT_WAITING → unlocks [C1]
[B1] "hg commit/pull" is started
[B2] "hg commit/pull" is ready to be committed
[B3] "hg commit/pull" spawn "internal" using a pretxnclose hook (need [C4])
[C1] "internal" waits on EXT_WAITING (need [A2])
[C2] "internal" creates EXT_UNLOCK → unlocks [A2]
[C3] "internal" show the tipmost revision (inside of the transaction)
[C4] "internal" waits on EXT_DONE (need [A4])
[A3] "external" show the tipmost revision (outside of the transaction)
[A4] "external" creates EXT_DONE → unlocks [C4]
[C5] "internal" end of execution -> unlock [B3]
[B4] "hg commit/pull" transaction is committed on disk
$ mkdir sync
$ mkdir output
$ mkdir script
$ HG_TEST_FILE_EXT_WAITING=$TESTTMP/sync/ext_waiting
$ export HG_TEST_FILE_EXT_WAITING
$ HG_TEST_FILE_EXT_UNLOCK=$TESTTMP/sync/ext_unlock
$ export HG_TEST_FILE_EXT_UNLOCK
$ HG_TEST_FILE_EXT_DONE=$TESTTMP/sync/ext_done
$ export HG_TEST_FILE_EXT_DONE
$ cat << EOF > script/external.sh
> #!/bin/sh
> "$RUNTESTDIR/testlib/wait-on-file" 5 "$HG_TEST_FILE_EXT_UNLOCK" "$HG_TEST_FILE_EXT_WAITING"
> hg log --rev 'tip' -T 'external: {rev} {desc}\n' > "$TESTTMP/output/external.out"
> touch "$HG_TEST_FILE_EXT_DONE"
> EOF
$ cat << EOF > script/internal.sh
> #!/bin/sh
> "$RUNTESTDIR/testlib/wait-on-file" 5 "$HG_TEST_FILE_EXT_WAITING"
> touch "$HG_TEST_FILE_EXT_UNLOCK"
> hg log --rev 'tip' -T 'internal: {rev} {desc}\n' > "$TESTTMP/output/internal.out"
> "$RUNTESTDIR/testlib/wait-on-file" 5 "$HG_TEST_FILE_EXT_DONE"
> EOF
Automated commands:
$ make_one_commit() {
> rm -f $TESTTMP/sync/*
> rm -f $TESTTMP/output/*
> hg log --rev 'tip' -T 'pre-commit: {rev} {desc}\n'
> echo x >> a
> sh $TESTTMP/script/external.sh & hg commit -m "$1"
> cat $TESTTMP/output/external.out
> cat $TESTTMP/output/internal.out
> hg log --rev 'tip' -T 'post-tr: {rev} {desc}\n'
> }
$ make_one_pull() {
> rm -f $TESTTMP/sync/*
> rm -f $TESTTMP/output/*
> hg log --rev 'tip' -T 'pre-commit: {rev} {desc}\n'
> echo x >> a
> sh $TESTTMP/script/external.sh & hg pull ../other-repo/ --rev "$1" --force --quiet
> cat $TESTTMP/output/external.out
> cat $TESTTMP/output/internal.out
> hg log --rev 'tip' -T 'post-tr: {rev} {desc}\n'
> }
prepare a large source to which to pull from:
The source is large to unsure we don't use inline more after the pull
$ hg init other-repo
$ hg -R other-repo debugbuilddag .+500
prepare an empty repository where to make test:
$ hg init repo
$ cd repo
$ touch a
$ hg add a
prepare a small extension to controll inline size
$ mkdir $TESTTMP/ext
$ cat << EOF > $TESTTMP/ext/small_inline.py
> from mercurial import revlog
> revlog._maxinline = 64 * 100
> EOF
$ cat << EOF >> $HGRCPATH
> [extensions]
> small_inline=$TESTTMP/ext/small_inline.py
> [hooks]
> pretxnclose = sh $TESTTMP/script/internal.sh
> EOF
check this is true for the initial commit (inline → inline)
-----------------------------------------------------------
the repository should still be inline (for relevant format)
$ make_one_commit first
pre-commit: -1
external: -1
internal: 0 first
post-tr: 0 first
#if revlogv1
$ hg debugrevlog -c | grep inline
flags : inline
#endif
check this is true for extra commit (inline → inline)
-----------------------------------------------------
the repository should still be inline (for relevant format)
#if revlogv1
$ hg debugrevlog -c | grep inline
flags : inline
#endif
$ make_one_commit second
pre-commit: 0 first
external: 0 first
internal: 1 second
post-tr: 1 second
#if revlogv1
$ hg debugrevlog -c | grep inline
flags : inline
#endif
check this is true for a small pull (inline → inline)
-----------------------------------------------------
the repository should still be inline (for relevant format)
#if revlogv1
$ hg debugrevlog -c | grep inline
flags : inline
#endif
$ make_one_pull 3
pre-commit: 1 second
warning: repository is unrelated
external: 1 second
internal: 5 r3
post-tr: 5 r3
#if revlogv1
$ hg debugrevlog -c | grep inline
flags : inline
#endif
Make a large pull (inline → no-inline)
---------------------------------------
the repository should no longer be inline (for relevant format)
#if revlogv1
$ hg debugrevlog -c | grep inline
flags : inline
#endif
$ make_one_pull 400
pre-commit: 5 r3
external: 5 r3
internal: 402 r400
post-tr: 402 r400
#if revlogv1
$ hg debugrevlog -c | grep inline
[1]
#endif
check this is true for extra commit (no-inline → no-inline)
-----------------------------------------------------------
the repository should no longer be inline (for relevant format)
#if revlogv1
$ hg debugrevlog -c | grep inline
[1]
#endif
$ make_one_commit third
pre-commit: 402 r400
external: 402 r400
internal: 403 third
post-tr: 403 third
#if revlogv1
$ hg debugrevlog -c | grep inline
[1]
#endif
Make a pull (not-inline → no-inline)
-------------------------------------
the repository should no longer be inline (for relevant format)
#if revlogv1
$ hg debugrevlog -c | grep inline
[1]
#endif
$ make_one_pull tip
pre-commit: 403 third
external: 403 third
internal: 503 r500
post-tr: 503 r500
#if revlogv1
$ hg debugrevlog -c | grep inline
[1]
#endif