revlog: automatically read from opened file handles
The revlog reading code commonly opens a new file handle for
reading on demand. There is support for passing a file handle
to revlog.revision(). But it is marked as an internal argument.
When revlogs are written, we write() data as it is available. But
we don't flush() data until all revisions are written.
Putting these two traits together, it is possible for an in-process
revlog reader during active writes to trigger the opening of a new
file handle on a file with unflushed writes. The reader won't have
access to all "available" revlog data (as it hasn't been flushed).
And with the introduction of the previous patch, this can lead to
the revlog raising an error due to a partial read.
I witnessed this behavior when applying changegroup data (via
`hg pull`) before issue6006 was fixed via different means. Having
this and the previous patch in play would have helped cause errors
earlier rather than manifesting as hash verification failures.
While this has been a long-standing issue, I believe the relatively
new delta computation code has tickled it into being more common.
This is because the new delta computation code will compute deltas
in more scenarios. This can lead to revlog reading. While the delta
computation code is probably supposed to reuse file handles, it
appears it isn't doing so in all circumstances.
But the issue runs deeper than that. Theoretically, any code can
access revision data during revlog writes. It appears we were just
getting lucky that it wasn't. (The "add revision callback" passed to
addgroup() provides an avenue to do this.)
If I changed the revlog's behavior to not cache the full revision
text or to clear caches after revision insertion during addgroup(),
I was able to produce crashes 100% of the time when writing changelog
revisions. This is because changelog's add revision callback attempts
to resolve the revision data to access the changed files list. And
without the revision's fulltext being cached, we performed a revlog
read, which required opening a new file handle. This attempted to read
unflushed data, leading to a partial read and a crash.
This commit teaches the revlog to store the file handles used for
writing multiple revisions during addgroup(). It also teaches the
code for resolving a file handle when reading to use these handles,
if available. This ensures that *any* reads (regardless of their
source) use the active writing file handles, if available. These
file handles have access to the unflushed data because they wrote it.
This allows reads to complete without issue.
Differential Revision: https://phab.mercurial-scm.org/D5267
$ cat <<EOF >> $HGRCPATH
> [extensions]
> mq =
> [diff]
> nodates = 1
> EOF
$ hg init a
$ cd a
$ mkdir 1 2
$ echo 'base' > 1/base
$ echo 'base' > 2/base
$ hg ci -Ambase
adding 1/base
adding 2/base
$ hg qnew -mmqbase mqbase
$ echo 'patched' > 1/base
$ echo 'patched' > 2/base
$ hg qrefresh
$ hg qdiff
diff -r e7af5904b465 1/base
--- a/1/base
+++ b/1/base
@@ -1,1 +1,1 @@
-base
+patched
diff -r e7af5904b465 2/base
--- a/2/base
+++ b/2/base
@@ -1,1 +1,1 @@
-base
+patched
$ hg qdiff .
diff -r e7af5904b465 1/base
--- a/1/base
+++ b/1/base
@@ -1,1 +1,1 @@
-base
+patched
diff -r e7af5904b465 2/base
--- a/2/base
+++ b/2/base
@@ -1,1 +1,1 @@
-base
+patched
$ cat .hg/patches/mqbase
# HG changeset patch
# Parent e7af5904b465cd1f4f3cf6b26fe14e8db6f63eaa
mqbase
diff -r e7af5904b465 1/base
--- a/1/base
+++ b/1/base
@@ -1,1 +1,1 @@
-base
+patched
diff -r e7af5904b465 2/base
--- a/2/base
+++ b/2/base
@@ -1,1 +1,1 @@
-base
+patched
$ echo 'patched again' > base
$ hg qrefresh 1
$ hg qdiff
diff -r e7af5904b465 1/base
--- a/1/base
+++ b/1/base
@@ -1,1 +1,1 @@
-base
+patched
diff -r e7af5904b465 2/base
--- a/2/base
+++ b/2/base
@@ -1,1 +1,1 @@
-base
+patched
$ hg qdiff .
diff -r e7af5904b465 1/base
--- a/1/base
+++ b/1/base
@@ -1,1 +1,1 @@
-base
+patched
diff -r e7af5904b465 2/base
--- a/2/base
+++ b/2/base
@@ -1,1 +1,1 @@
-base
+patched
$ cat .hg/patches/mqbase
# HG changeset patch
# Parent e7af5904b465cd1f4f3cf6b26fe14e8db6f63eaa
mqbase
diff -r e7af5904b465 1/base
--- a/1/base
+++ b/1/base
@@ -1,1 +1,1 @@
-base
+patched
qrefresh . in subdir:
$ ( cd 1 ; hg qrefresh . )
$ hg qdiff
diff -r e7af5904b465 1/base
--- a/1/base
+++ b/1/base
@@ -1,1 +1,1 @@
-base
+patched
diff -r e7af5904b465 2/base
--- a/2/base
+++ b/2/base
@@ -1,1 +1,1 @@
-base
+patched
$ hg qdiff .
diff -r e7af5904b465 1/base
--- a/1/base
+++ b/1/base
@@ -1,1 +1,1 @@
-base
+patched
diff -r e7af5904b465 2/base
--- a/2/base
+++ b/2/base
@@ -1,1 +1,1 @@
-base
+patched
$ cat .hg/patches/mqbase
# HG changeset patch
# Parent e7af5904b465cd1f4f3cf6b26fe14e8db6f63eaa
mqbase
diff -r e7af5904b465 1/base
--- a/1/base
+++ b/1/base
@@ -1,1 +1,1 @@
-base
+patched
qrefresh in hg-root again:
$ hg qrefresh
$ hg qdiff
diff -r e7af5904b465 1/base
--- a/1/base
+++ b/1/base
@@ -1,1 +1,1 @@
-base
+patched
diff -r e7af5904b465 2/base
--- a/2/base
+++ b/2/base
@@ -1,1 +1,1 @@
-base
+patched
$ hg qdiff .
diff -r e7af5904b465 1/base
--- a/1/base
+++ b/1/base
@@ -1,1 +1,1 @@
-base
+patched
diff -r e7af5904b465 2/base
--- a/2/base
+++ b/2/base
@@ -1,1 +1,1 @@
-base
+patched
$ cat .hg/patches/mqbase
# HG changeset patch
# Parent e7af5904b465cd1f4f3cf6b26fe14e8db6f63eaa
mqbase
diff -r e7af5904b465 1/base
--- a/1/base
+++ b/1/base
@@ -1,1 +1,1 @@
-base
+patched
diff -r e7af5904b465 2/base
--- a/2/base
+++ b/2/base
@@ -1,1 +1,1 @@
-base
+patched
qrefresh --short tests:
$ echo 'orphan' > orphanchild
$ hg add orphanchild
$ hg qrefresh nonexistentfilename # clear patch
nonexistentfilename: * (glob)
$ hg diff -c qtip
$ hg qrefresh --short 1/base
$ hg qrefresh --short 2/base
$ hg qdiff
diff -r e7af5904b465 1/base
--- a/1/base
+++ b/1/base
@@ -1,1 +1,1 @@
-base
+patched
diff -r e7af5904b465 2/base
--- a/2/base
+++ b/2/base
@@ -1,1 +1,1 @@
-base
+patched
diff -r e7af5904b465 orphanchild
--- /dev/null
+++ b/orphanchild
@@ -0,0 +1,1 @@
+orphan
$ cat .hg/patches/mqbase
# HG changeset patch
# Parent e7af5904b465cd1f4f3cf6b26fe14e8db6f63eaa
mqbase
diff -r e7af5904b465 1/base
--- a/1/base
+++ b/1/base
@@ -1,1 +1,1 @@
-base
+patched
diff -r e7af5904b465 2/base
--- a/2/base
+++ b/2/base
@@ -1,1 +1,1 @@
-base
+patched
$ hg st
A orphanchild
? base
diff shows what is not in patch:
$ hg diff
diff -r ???????????? orphanchild (glob)
--- /dev/null
+++ b/orphanchild
@@ -0,0 +1,1 @@
+orphan
Before starting exclusive tests:
$ cat .hg/patches/mqbase
# HG changeset patch
# Parent e7af5904b465cd1f4f3cf6b26fe14e8db6f63eaa
mqbase
diff -r e7af5904b465 1/base
--- a/1/base
+++ b/1/base
@@ -1,1 +1,1 @@
-base
+patched
diff -r e7af5904b465 2/base
--- a/2/base
+++ b/2/base
@@ -1,1 +1,1 @@
-base
+patched
Exclude 2/base:
$ hg qref -s -X 2/base
$ cat .hg/patches/mqbase
# HG changeset patch
# Parent e7af5904b465cd1f4f3cf6b26fe14e8db6f63eaa
mqbase
diff -r e7af5904b465 1/base
--- a/1/base
+++ b/1/base
@@ -1,1 +1,1 @@
-base
+patched
status shows 2/base as dirty:
$ hg status
M 2/base
A orphanchild
? base
Remove 1/base and add 2/base again but not orphanchild:
$ hg qref -s -X orphanchild -X 1/base 2/base orphanchild
$ cat .hg/patches/mqbase
# HG changeset patch
# Parent e7af5904b465cd1f4f3cf6b26fe14e8db6f63eaa
mqbase
diff -r e7af5904b465 2/base
--- a/2/base
+++ b/2/base
@@ -1,1 +1,1 @@
-base
+patched
Add 1/base with include filter - and thus remove 2/base from patch:
$ hg qref -s -I 1/ o* */*
$ cat .hg/patches/mqbase
# HG changeset patch
# Parent e7af5904b465cd1f4f3cf6b26fe14e8db6f63eaa
mqbase
diff -r e7af5904b465 1/base
--- a/1/base
+++ b/1/base
@@ -1,1 +1,1 @@
-base
+patched
$ cd ..
Test qrefresh --git losing copy metadata:
$ hg init repo
$ cd repo
$ echo "[diff]" >> .hg/hgrc
$ echo "git=True" >> .hg/hgrc
$ echo a > a
$ hg ci -Am adda
adding a
$ hg copy a ab
$ echo b >> ab
$ hg copy a ac
$ echo c >> ac
Capture changes:
$ hg qnew -f p1
$ hg qdiff
diff --git a/a b/ab
copy from a
copy to ab
--- a/a
+++ b/ab
@@ -1,1 +1,2 @@
a
+b
diff --git a/a b/ac
copy from a
copy to ac
--- a/a
+++ b/ac
@@ -1,1 +1,2 @@
a
+c
Refresh and check changes again:
$ hg qrefresh
$ hg qdiff
diff --git a/a b/ab
copy from a
copy to ab
--- a/a
+++ b/ab
@@ -1,1 +1,2 @@
a
+b
diff --git a/a b/ac
copy from a
copy to ac
--- a/a
+++ b/ac
@@ -1,1 +1,2 @@
a
+c
$ cd ..
Issue1441: qrefresh confused after hg rename:
$ hg init repo-1441
$ cd repo-1441
$ echo a > a
$ hg add a
$ hg qnew -f p
$ hg mv a b
$ hg qrefresh
$ hg qdiff
diff -r 000000000000 b
--- /dev/null
+++ b/b
@@ -0,0 +1,1 @@
+a
$ cd ..
Issue2025: qrefresh does not honor filtering options when tip !=
qtip:
$ hg init repo-2025
$ cd repo-2025
$ echo a > a
$ echo b > b
$ hg ci -qAm addab
$ echo a >> a
$ echo b >> b
$ hg qnew -f patch
$ hg up -qC 0
$ echo c > c
$ hg ci -qAm addc
$ hg up -qC 1
refresh with tip != qtip:
$ hg --config diff.nodates=1 qrefresh -I b
$ hg st
M a
$ cat b
b
b
$ cat .hg/patches/patch
# HG changeset patch
# Parent 1a60229be7ac3e4a7f647508e99b87bef1f03593
diff -r 1a60229be7ac b
--- a/b
+++ b/b
@@ -1,1 +1,2 @@
b
+b
$ cd ..
Issue1441 with git patches:
$ hg init repo-1441-git
$ cd repo-1441-git
$ echo "[diff]" >> .hg/hgrc
$ echo "git=True" >> .hg/hgrc
$ echo a > a
$ hg add a
$ hg qnew -f p
$ hg mv a b
$ hg qrefresh
$ hg qdiff --nodates
diff --git a/b b/b
new file mode 100644
--- /dev/null
+++ b/b
@@ -0,0 +1,1 @@
+a
$ cd ..
Refresh with bad usernames. Mercurial used to abort on bad usernames,
but only after writing the bad name into the patch.
$ hg init bad-usernames
$ cd bad-usernames
$ touch a
$ hg add a
$ hg qnew a
$ hg qrefresh -u 'foo
> bar'
transaction abort!
rollback completed
qrefresh interrupted while patch was popped! (revert --all, qpush to recover)
abort: username 'foo\nbar' contains a newline!
[255]
$ rm a
$ cat .hg/patches/a
# HG changeset patch
# Parent 0000000000000000000000000000000000000000
diff --git a/a b/a
new file mode 100644
$ hg qpush
applying a
now at: a
$ hg qrefresh -u ' '
transaction abort!
rollback completed
qrefresh interrupted while patch was popped! (revert --all, qpush to recover)
abort: empty username!
[255]
$ cat .hg/patches/a
# HG changeset patch
# Parent 0000000000000000000000000000000000000000
diff --git a/a b/a
new file mode 100644
$ cd ..
Refresh with phase data:
$ cd repo
$ echo 'babar' >> a
$ hg qnew -m 'update a' p2.diff
$ hg phase p2.diff
2: draft
$ echo 'beber' >> a
$ hg qref
$ hg phase p2.diff
2: draft
$ hg phase --force --secret p2.diff
$ echo 'bibir' >> a
$ hg qref
$ hg phase p2.diff
2: secret
$ cd ..