1
Detecting unchanged files
2
*************************
4
Many operations need to know which working files have changed compared
5
to the basis revision. (We also sometimes want to know which files
6
have changed between two revisions, but since we know the text-ids
7
and hashes that is much easier.)
9
The simplest way is to just directly compare the files. This is
10
simple and reliable, but has the disadvantage that we need to read in
11
both files. For a large tree like the kernel or even samba, this can
12
use a lot of cache memory and/or be slow. Some people have machines
13
that do not have enough memory to hold even one copy of the tree at a
14
time, and this would use two copies.
16
So it is nice to know which files have not changed without actually
17
reading them. Possibilities:
19
* Make the working files be hardlinks to the file store. Easy to see
20
if they are still the same file or not by simply stat'ing them. For
21
extra protection, make the stored files readonly. Has the
22
additional advantage of reducing the disk usage of the working copy.
24
Disadvantage is that some people have editors that do not handle
25
this safely. In that case the changes will go undetected, and they
26
could corrupt history. Pretty bad.
28
We can provide a ``bzr edit`` command that breaks the link and makes
29
the working copy writable.
31
* As above, but link to a temporary pristine directory, not to the
32
real store. They can get a wrong answer, but at least cannot
35
* Check the mtime and size of the file; compare them to those of the
36
previous stored version.
38
The mtime doesn't need to be the time the previous revision was
41
There is a possibility of a race here where the file is modified but
42
does not change size, all in the second after the checkout. Many
43
filesystems don't report sub-second modification times, but Linux
44
now allows for it and it may be supported in future.
46
* Read in all the working files, but first compare them to the
47
size and text-hash of the previous file revision; only do the diff
48
if they have actually changed. Means only reading one tree, not
49
two, but we still have to scan all the source.
51
* Copy the file, but use an explicit edit command to remember which
52
ones have been changed. Uneditable files should be readonly to
53
prevent inadvertent changes.
55
The problem with almost all of these models is that it is possible for
56
people to change a file without tripping them. The only way to make
57
this perfectly safe is to actually compare. So perhaps there should
60
It is crucial that no failure can lose history. Does that mean
61
hardlinks directly into the file store are just too risky?
63
It is most important that ``diff``, ``status`` and similar things be
64
fast, because they are invoked often. It may be that the ``commit``
65
command can tolerate being somewhat slower -- but then it would be
66
confusing if ``commit`` saw something different to what ``diff`` does,
67
so they should be the same.
69
For the mooted `edit command`__, we will know whether a file is
70
checked out as r/o or r/w; if a file is known to be read-only it can
71
be assumed to be unmodified.
75
The ``check`` command can make sure that any files that would be
76
assumed to be unmodified are actually unmodified.
78
File times can skew, particularly on network filesystems. We should
79
not make any assumptions that mtimes are the same as the system clock
80
at the time of creation (and that would probably be racy anyhow).
87
if file is not editable:
91
if same inum and device as basis:
93
elif present as read-only:
95
elif same mtime and size:
98
read working file, calculate sha1
99
if same size and sha-1 as previous inventory-entry:
added
removed
doc/unchanged.txt
