Mercurial > hg
view tests/test-diff-unified.t @ 46607:e9901d01d135
revlog: add a mechanism to verify expected file position before appending
If someone uses `hg debuglocks`, or some non-hg process writes to the .hg
directory without respecting the locks, or if the repo's on a networked
filesystem, it's possible for the revlog code to write out corrupted data.
The form of this corruption can vary depending on what data was written and how
that happened. We are in the "networked filesystem" case (though I've had users
also do this to themselves with the "`hg debuglocks`" scenario), and most often
see this with the changelog. What ends up happening is we produce two items
(let's call them rev1 and rev2) in the .i file that have the same linkrev,
baserev, and offset into the .d file, while the data in the .d file is appended
properly. rev2's compressed_size is accurate for rev2, but when we go to
decompress the data in the .d file, we use the offset that's recorded in the
index file, which is the same as rev1, and attempt to decompress
rev2.compressed_size bytes of rev1's data. This usually does not succeed. :)
When using inline data, this also fails, though I haven't investigated why too
closely. This shows up as a "patch decode" error. I believe what's happening
there is that we're basically ignoring the offset field, getting the data
properly, but since baserev != rev, it thinks this is a delta based on rev
(instead of a full text) and can't actually apply it as such.
For now, I'm going to make this an optional component and default it to entirely
off. I may increase the default severity of this in the future, once I've
enabled it for my users and we gain more experience with it. Luckily, most of my
users have a versioned filesystem and can roll back to before the corruption has
been written, it's just a hassle to do so and not everyone knows how (so it's a
support burden). Users on other filesystems will not have that luxury, and this
can cause them to have a corrupted repository that they are unlikely to know how
to resolve, and they'll see this as a data-loss event. Refusing to create the
corruption is a much better user experience.
This mechanism is not perfect. There may be false-negatives (racy writes that
are not detected). There should not be any false-positives (non-racy writes that
are detected as such). This is not a mechanism that makes putting a repo on a
networked filesystem "safe" or "supported", just *less* likely to cause
corruption.
Differential Revision: https://phab.mercurial-scm.org/D9952
author | Kyle Lippincott <spectral@google.com> |
---|---|
date | Wed, 03 Feb 2021 16:33:10 -0800 |
parents | 22f19da5736e |
children | df44bac9413d |
line wrap: on
line source
$ hg init repo $ cd repo $ cat > a <<EOF > c > c > a > a > b > a > a > c > c > EOF $ hg ci -Am adda adding a $ cat > a <<EOF > c > c > a > a > dd > a > a > c > c > EOF default context $ hg diff --nodates diff -r cf9f4ba66af2 a --- a/a +++ b/a @@ -2,7 +2,7 @@ c a a -b +dd a a c invalid --unified $ hg diff --nodates -U foo abort: diff context lines count must be an integer, not 'foo' [255] $ hg diff --nodates -U 2 diff -r cf9f4ba66af2 a --- a/a +++ b/a @@ -3,5 +3,5 @@ a a -b +dd a a $ hg --config diff.unified=2 diff --nodates diff -r cf9f4ba66af2 a --- a/a +++ b/a @@ -3,5 +3,5 @@ a a -b +dd a a $ hg diff --nodates -U 1 diff -r cf9f4ba66af2 a --- a/a +++ b/a @@ -4,3 +4,3 @@ a -b +dd a invalid diff.unified $ hg --config diff.unified=foo diff --nodates abort: diff context lines count must be an integer, not 'foo' [255] noprefix config and option $ hg --config diff.noprefix=True diff --nodates diff -r cf9f4ba66af2 a --- a +++ a @@ -2,7 +2,7 @@ c a a -b +dd a a c $ hg diff --noprefix --nodates diff -r cf9f4ba66af2 a --- a +++ a @@ -2,7 +2,7 @@ c a a -b +dd a a c noprefix config disabled in plain mode, but option still enabled $ HGPLAIN=1 hg --config diff.noprefix=True diff --nodates diff -r cf9f4ba66af2 a --- a/a +++ b/a @@ -2,7 +2,7 @@ c a a -b +dd a a c $ HGPLAIN=1 hg diff --noprefix --nodates diff -r cf9f4ba66af2 a --- a +++ a @@ -2,7 +2,7 @@ c a a -b +dd a a c $ cd .. 0 lines of context hunk header matches gnu diff hunk header $ hg init diffzero $ cd diffzero $ cat > f1 << EOF > c2 > c4 > c5 > EOF $ hg commit -Am0 adding f1 $ cat > f2 << EOF > c1 > c2 > c3 > c4 > EOF $ mv f2 f1 $ hg diff -U0 --nodates diff -r 55d8ff78db23 f1 --- a/f1 +++ b/f1 @@ -0,0 +1,1 @@ +c1 @@ -1,0 +3,1 @@ +c3 @@ -3,1 +4,0 @@ -c5 $ hg diff -U0 --nodates --git diff --git a/f1 b/f1 --- a/f1 +++ b/f1 @@ -0,0 +1,1 @@ +c1 @@ -1,0 +3,1 @@ +c3 @@ -3,1 +4,0 @@ -c5 $ hg diff -U0 --nodates -p diff -r 55d8ff78db23 f1 --- a/f1 +++ b/f1 @@ -0,0 +1,1 @@ +c1 @@ -1,0 +3,1 @@ c2 +c3 @@ -3,1 +4,0 @@ c4 -c5 $ echo a > f1 $ hg ci -m movef2 Test diff headers terminating with TAB when necessary (issue3357) Regular diff --nodates, file creation $ hg mv f1 'f 1' $ echo b > 'f 1' $ hg diff --nodates 'f 1' diff -r 7574207d0d15 f 1 --- /dev/null +++ b/f 1 @@ -0,0 +1,1 @@ +b Git diff, adding space $ hg diff --git diff --git a/f1 b/f 1 rename from f1 rename to f 1 --- a/f1 +++ b/f 1 @@ -1,1 +1,1 @@ -a +b Git diff, adding extended headers $ hg diff --git --config experimental.extendedheader.index=7 --config experimental.extendedheader.similarity=True diff --git a/f1 b/f 1 similarity index 0% rename from f1 rename to f 1 index 7898192..6178079 100644 --- a/f1 +++ b/f 1 @@ -1,1 +1,1 @@ -a +b $ hg diff --git --config experimental.extendedheader.index=-1 invalid length for extendedheader.index: '-1' diff --git a/f1 b/f 1 rename from f1 rename to f 1 --- a/f1 +++ b/f 1 @@ -1,1 +1,1 @@ -a +b $ hg diff --git --config experimental.extendedheader.index=whatever invalid value for extendedheader.index: 'whatever' diff --git a/f1 b/f 1 rename from f1 rename to f 1 --- a/f1 +++ b/f 1 @@ -1,1 +1,1 @@ -a +b Git diff with noprefix $ hg --config diff.noprefix=True diff --git --nodates diff --git f1 f 1 rename from f1 rename to f 1 --- f1 +++ f 1 @@ -1,1 +1,1 @@ -a +b noprefix config disabled in plain mode, but option still enabled $ HGPLAIN=1 hg --config diff.noprefix=True diff --git --nodates diff --git a/f1 b/f 1 rename from f1 rename to f 1 --- a/f1 +++ b/f 1 @@ -1,1 +1,1 @@ -a +b $ HGPLAIN=1 hg diff --git --noprefix --nodates diff --git f1 f 1 rename from f1 rename to f 1 --- f1 +++ f 1 @@ -1,1 +1,1 @@ -a +b Regular diff --nodates, file deletion $ hg ci -m addspace $ hg mv 'f 1' f1 $ echo a > f1 $ hg diff --nodates 'f 1' diff -r ca50fe67c9c7 f 1 --- a/f 1 +++ /dev/null @@ -1,1 +0,0 @@ -b Git diff, removing space $ hg diff --git diff --git a/f 1 b/f1 rename from f 1 rename to f1 --- a/f 1 +++ b/f1 @@ -1,1 +1,1 @@ -b +a showfunc diff $ cat > f1 << EOF > int main() { > int a = 0; > int b = 1; > int c = 2; > int d = 3; > return a + b + c + d; > } > EOF $ hg commit -m addfunction $ cat > f1 << EOF > int main() { > int a = 0; > int b = 1; > int c = 2; > int e = 3; > return a + b + c + e; > } > EOF $ hg diff --git diff --git a/f1 b/f1 --- a/f1 +++ b/f1 @@ -2,6 +2,6 @@ int a = 0; int b = 1; int c = 2; - int d = 3; - return a + b + c + d; + int e = 3; + return a + b + c + e; } $ hg diff --config diff.showfunc=True --git diff --git a/f1 b/f1 --- a/f1 +++ b/f1 @@ -2,6 +2,6 @@ int main() { int a = 0; int b = 1; int c = 2; - int d = 3; - return a + b + c + d; + int e = 3; + return a + b + c + e; } If [diff] git is set to true, but the user says --no-git, we should *not* get git diffs $ hg diff --nodates --config diff.git=1 --no-git diff -r f2c7c817fa55 f1 --- a/f1 +++ b/f1 @@ -2,6 +2,6 @@ int a = 0; int b = 1; int c = 2; - int d = 3; - return a + b + c + d; + int e = 3; + return a + b + c + e; } $ cd .. Long function names should be abbreviated, but multi-byte character shouldn't be broken up $ hg init longfunc $ cd longfunc >>> with open('a', 'wb') as f: ... f.write(b'a' * 39 + b'bb' + b'\n') and None ... f.write(b' .\n' * 3) and None ... f.write(b' 0 b\n') and None ... f.write(b' .\n' * 3) and None ... f.write(b'a' * 39 + b'\xc3\xa0' + b'\n') and None ... f.write(b' .\n' * 3) and None ... f.write(b' 0 a with grave (single code point)\n') and None ... f.write(b' .\n' * 3) and None ... f.write(b'a' * 39 + b'a\xcc\x80' + b'\n') and None ... f.write(b' .\n' * 3) and None ... f.write(b' 0 a with grave (composition)\n') and None ... f.write(b' .\n' * 3) and None $ hg ci -qAm0 >>> with open('a', 'wb') as f: ... f.write(b'a' * 39 + b'bb' + b'\n') and None ... f.write(b' .\n' * 3) and None ... f.write(b' 1 b\n') and None ... f.write(b' .\n' * 3) and None ... f.write(b'a' * 39 + b'\xc3\xa0' + b'\n') and None ... f.write(b' .\n' * 3) and None ... f.write(b' 1 a with grave (single code point)\n') and None ... f.write(b' .\n' * 3) and None ... f.write(b'a' * 39 + b'a\xcc\x80' + b'\n') and None ... f.write(b' .\n' * 3) and None ... f.write(b' 1 a with grave (composition)\n') and None ... f.write(b' .\n' * 3) and None $ hg ci -m1 $ hg diff -c1 --nodates --show-function diff -r 3e92dd6fa812 -r a256341606cb a --- a/a +++ b/a @@ -2,7 +2,7 @@ aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaab . . . - 0 b + 1 b . . . @@ -10,7 +10,7 @@ aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa\xc3\xa0 (esc) . . . - 0 a with grave (single code point) + 1 a with grave (single code point) . . . @@ -18,7 +18,7 @@ aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa\xcc\x80 (esc) . . . - 0 a with grave (composition) + 1 a with grave (composition) . . . $ cd ..