When running a required clean filter, we do not have to mmap the
original before feeding the filter. Instead, stream the file
contents directly to the filter and process its output.
* sp/stream-clean-filter:
sha1_file: don't convert off_t to size_t too early to avoid potential die()
convert: stream from fd to required clean filter to reduce used address space
copy_fd(): do not close the input file descriptor
mmap_limit: introduce GIT_MMAP_LIMIT to allow testing expected mmap size
memory_limit: use git_env_ulong() to parse GIT_ALLOC_LIMIT
config.c: add git_env_ulong() to parse environment variable
convert: drop arguments other than 'path' from would_convert_to_git()
The data is streamed to the filter process anyway. Better avoid mapping
the file if possible. This is especially useful if a clean filter
reduces the size, for example if it computes a sha1 for binary data,
like git media. The file size that the previous implementation could
handle was limited by the available address space; large files for
example could not be handled with (32-bit) msysgit. The new
implementation can filter files of any size as long as the filter output
is small enough.
The new code path is only taken if the filter is required. The filter
consumes data directly from the fd. If it fails, the original data is
not immediately available. The condition can easily be handled as
a fatal error, which is expected for a required filter anyway.
If the filter was not required, the condition would need to be handled
in a different way, like seeking to 0 and reading the data. But this
would require more restructuring of the code and is probably not worth
it. The obvious approach of falling back to reading all data would not
help achieving the main purpose of this patch, which is to handle large
files with limited address space. If reading all data is an option, we
can simply take the old code path right away and mmap the entire file.
The environment variable GIT_MMAP_LIMIT, which has been introduced in
a previous commit is used to test that the expected code path is taken.
A related test that exercises required filters is modified to verify
that the data actually has been modified on its way from the file system
to the object store.
Signed-off-by: Steffen Prohaska <prohaska@zib.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Most struct child_process variables are cleared using memset first after
declaration. Provide a macro, CHILD_PROCESS_INIT, that can be used to
initialize them statically instead. That's shorter, doesn't require a
function call and is slightly more readable (especially given that we
already have STRBUF_INIT, ARGV_ARRAY_INIT etc.).
Helped-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Rene Scharfe <l.s.r@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
It's a common idiom to match a prefix and then skip past it
with a magic number, like:
if (starts_with(foo, "bar"))
foo += 3;
This is easy to get wrong, since you have to count the
prefix string yourself, and there's no compiler check if the
string changes. We can use skip_prefix to avoid the magic
numbers here.
Note that some of these conversions could be much shorter.
For example:
if (starts_with(arg, "--foo=")) {
bar = arg + 6;
continue;
}
could become:
if (skip_prefix(arg, "--foo=", &bar))
continue;
However, I have left it as:
if (skip_prefix(arg, "--foo=", &v)) {
bar = v;
continue;
}
to visually match nearby cases which need to actually
process the string. Like:
if (skip_prefix(arg, "--foo=", &v)) {
bar = atoi(v);
continue;
}
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Leaving only the function definitions and declarations so that any
new topic in flight can still make use of the old functions, replace
existing uses of the prefixcmp() and suffixcmp() with new API
functions.
The change can be recreated by mechanically applying this:
$ git grep -l -e prefixcmp -e suffixcmp -- \*.c |
grep -v strbuf\\.c |
xargs perl -pi -e '
s|!prefixcmp\(|starts_with\(|g;
s|prefixcmp\(|!starts_with\(|g;
s|!suffixcmp\(|ends_with\(|g;
s|suffixcmp\(|!ends_with\(|g;
'
on the result of preparatory changes in this series.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Reduce duplicated code between convert.c and attr.c.
* lf/read-blob-data-from-index:
convert.c: remove duplicate code
read_blob_data_from_index(): optionally return the size of blob data
attr.c: extract read_index_data() as read_blob_data_from_index()
The has_cr_in_index() function is an almost 1:1 copy of
read_blob_data_from_index() with some additions. Use the
latter instead of using copy-pasted code.
Signed-off-by: Lukas Fleischer <git@cryptocrack.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
These callers can drop some inline pointer arithmetic and
magic offset constants, making them more readable and less
error-prone (those constants had to match the lengths of
strings, but there is no automatic verification of that
fact).
The "ep" pointer (presumably for "end pointer"), which
points to the final key segment of the config variable, is
given the more standard name "key" to describe its function
rather than its derivation.
Signed-off-by: Jeff King <peff@peff.net>
Reviewed-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* jk/maint-avoid-streaming-filtered-contents:
do not stream large files to pack when filters are in use
teach dry-run convert_to_git not to require a src buffer
teach convert_to_git a "dry run" mode
When we call convert_to_git in dry-run mode, it may still
want to look at the source buffer, because some CRLF
conversion modes depend on analyzing the source to determine
whether it is in fact convertible CRLF text.
However, the main motivation for convert_to_git's dry-run
mode is that we would decide which method to use to acquire
the blob's data (streaming versus in-core). Requiring this
source analysis creates a chicken-and-egg problem. We are
better off simply guessing that anything we can't analyze
will end up needing conversion.
This patch lets a caller specify a NULL src buffer when
using dry-run mode (and only dry-run mode). A non-zero
return value goes from "we would convert" to "we might
convert"; a zero return value remains "we would definitely
not convert".
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Some callers may want to know whether convert_to_git will
actually do anything before performing the conversion
itself (e.g., to decide whether to stream or handle blobs
in-core). This patch lets callers specify the dry run mode
by passing a NULL destination buffer. The return value,
instead of indicating whether conversion happened, will
indicate whether conversion would occur.
For readability, we also include a wrapper function which
makes it more obvious we are not actually performing the
conversion.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If a filter is not defined or if it fails, git should behave as if the
filter is a no-op passthru.
However, if the filter exits before reading all the content, depending on
the timing, git could be killed with SIGPIPE when it tries to write to the
pipe connected to the filter.
Ignore SIGPIPE while processing the filter to give us a chance to check
the return value from a failed write, in order to detect and act on this
mode of failure in a more controlled way.
Signed-off-by: Jehan Bing <jehan@orb.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
By default, a missing filter driver or a failure from the filter driver is
not an error, but merely makes the filter operation a no-op pass through.
This is useful to massage the content into a shape that is more convenient
for the platform, filesystem, and the user to use, and the content filter
mechanism is not used to turn something unusable into usable.
However, we could also use of the content filtering mechanism and store
the content that cannot be directly used in the repository (e.g. a UUID
that refers to the true content stored outside git, or an encrypted
content) and turn it into a usable form upon checkout (e.g. download the
external content, or decrypt the encrypted content). For such a use case,
the content cannot be used when filter driver fails, and we need a way to
tell Git to abort the whole operation for such a failing or missing filter
driver.
Add a new "filter.<driver>.required" configuration variable to mark the
second use case. When it is set, git will abort the operation when the
filter driver does not exist or exits with a non-zero status code.
Signed-off-by: Jehan Bing <jehan@orb.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The non-streaming version of the filter counts CRLF and LF in the whole
buffer, and returns without doing anything when they match (i.e. what is
recorded in the object store already uses CRLF). This was done to help
people who added files from the DOS world before realizing they want to go
cross platform and adding .gitattributes to tell Git that they only want
CRLF in their working tree.
The streaming version of the filter does not want to read the whole thing
before starting to work, as that defeats the whole point of streaming. So
we instead check what byte follows CR whenever we see one, and add CR
before LF only when the LF does not immediately follow CR already to keep
CRLF as is.
Reported-and-tested-by: Ralf Thielow
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This can only happen when the input size is multiple of the
buffer size of the cascade filter (16k) and ends with an LF,
but in such a case, the code forgot to tell the caller that
it added the "\n" it could not add during the last round.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
There may not be enough space to store CRLF in the output. If we don't
fill the buffer, then the filter will keep getting called with the same
short buffer and will loop forever.
Instead, always store the CR and record whether there's a missing LF
if so we store it in the output buffer the next time the function gets
called.
Reported-by: Henrik Grubbström <grubba@roxen.com>
Signed-off-by: Carlos Martín Nieto <cmn@elego.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The git_path_check_eol() function converts a string value to the
corresponding 'enum eol' value. However, the function is currently
declared to return an 'enum crlf_action', which causes sparse to
complain thus:
SP convert.c
convert.c:736:50: warning: mixing different enum types
convert.c:736:50: int enum crlf_action versus
convert.c:736:50: int enum eol
In order to suppress the warning, we simply correct the return type
in the function declaration.
Signed-off-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
convert_to_git sets src=dst->buf if any of the preceding conversions
actually did any work. Thus in ident_to_git we have to use memmove
instead of memcpy as far as src->dst copying is concerned.
Signed-off-by: Thomas Rast <trast@student.ethz.ch>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Suggested by: Junio Hamano <gitster@pobox.com>
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add support for "ident" filter on the output codepath. This does not work
with lf-to-crlf filter together (yet).
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Some filters may need to buffer the input and look-ahead inside it
to decide what to output, and they may consume more than zero bytes
of input and still not produce any output. After feeding all the
input, pass NULL as input as keep calling stream_filter() to let
such filters know there is no more input coming, and it is time for
them to produce the remaining output based on the buffered input.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This introduces an API to plug custom filters to an input stream.
The caller gets get_stream_filter("path") to obtain an appropriate
filter for the path, and then uses it when opening an input stream
via open_istream(). After that, the caller can read from the stream
with read_istream(), and close it with close_istream(), just like an
unfiltered stream.
This only adds a "null" filter that is a pass-thru filter, but later
changes can add LF-to-CRLF and other filters, and the callers of the
streaming API do not have to change.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When the output to a path does not have to be converted, we can read from
the object database from the streaming API and write to the file in the
working tree, without having to hold everything in the memory.
The ident, auto- and safe- crlf conversions inherently require you to read
the whole thing before deciding what to do, so while it is technically
possible to support them by using a buffer of an unbound size or rewinding
and reading the stream twice, it is less practical than the traditional
"read the whole thing in core and convert" approach.
Adding streaming filters for the other conversions on top of this should
be doable by tweaking the can_bypass_conversion() function (it should be
renamed to can_filter_stream() when it happens). Then the streaming API
can be extended to wrap the git_istream streaming_write_entry() opens on
the underlying object in another git_istream that reads from it, filters
what is read, and let the streaming_write_entry() read the filtered
result. But that is outside the scope of this series.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The current internal API requires the callers of setup_convert_check() to
supply the git_attr_check structures (hence they need to know how many to
allocate), but they grab the same set of attributes for given path.
Define a new convert_attrs() API that fills a higher level information that
the callers (convert_to_git and convert_to_working_tree) really want, and
move the common code to interact with the attributes system to it.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The places that need to pass an array of "struct git_attr_check" needed to
be careful to pass a large enough array and know what index each element
lied. Make it safer and easier to code these.
Besides, the hard-coded sequence of initializing various attributes was
too ugly after we gained more than a few attributes.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Back when the conversion was only about the end-of-line convention, it
might have made sense to call what we do upon seeing CR/LF simply an
"action", but these days the conversion routines do a lot more than just
tweaking the line ending. Raname "action" to "crlf_action".
The function that decides what end of line conversion to use on the output
codepath was called "determine_output_conversion", as if there is no other
kind of output conversion. Rename it to "output_eol"; it is a function
that returns what EOL convention is to be used.
A function that decides what "crlf_action" needs to be used on the input
codepath, given what conversion attribute is set to the path and global
end-of-line convention, was called "determine_action". Rename it to
"input_crlf_action".
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Yes, it is clear that "eol" wants to mean some sort of end-of-line thing,
but as the name of a global variable, it is way too short to describe what
kind of end-of-line thing it wants to represent. Besides, there are many
codepaths that want to use their own local "char *eol" variable to point
at the end of the current line they are processing.
This global variable holds what we read from core.eol configuration
variable. Name it as such.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since v1.7.2-rc0~23^2~2 (Add per-repository eol normalization,
2010-05-19), building with gcc -std=gnu89 -pedantic produces warnings
like the following:
convert.c:21:11: warning: comma at end of enumerator list [-pedantic]
gcc is right to complain --- these commas are not permitted in C89.
In the spirit of v1.7.2-rc0~32^2~16 (2010-05-14), remove them.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Filtering to support keyword expansion may need the name of
the file being filtered. In particular, to support p4 keywords
like
$File: //depot/product/dir/script.sh $
the smudge filter needs to know the name of the file it is
smudging.
Allow "%f" in the custom filter command line specified in the
configuration. This will be substituted by the filename
inside a single-quote pair to be passed to the shell.
Signed-off-by: Pete Wyckoff <pw@padd.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Disable CRLF expansion when convert_to_working_tree() is called from
normalize_buffer(). This improves performance when merging branches
with conflicting line endings when core.eol=crlf or core.autocrlf=true
by making the normalization act as if core.eol=lf.
Signed-off-by: Eyvind Bernhardsen <eyvind.bernhardsen@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Currently, merging across changes in line ending normalization is
painful since files containing CRLF will conflict with normalized files,
even if the only difference between the two versions is the line
endings. Additionally, any "real" merge conflicts that exist are
obscured because every line in the file has a conflict.
Assume you start out with a repo that has a lot of text files with CRLF
checked in (A):
o---C
/ \
A---B---D
B: Add "* text=auto" to .gitattributes and normalize all files to
LF-only
C: Modify some of the text files
D: Try to merge C
You will get a ridiculous number of LF/CRLF conflicts when trying to
merge C into D, since the repository contents for C are "wrong" wrt the
new .gitattributes file.
Fix ll-merge so that the "base", "theirs" and "ours" stages are passed
through convert_to_worktree() and convert_to_git() before a three-way
merge. This ensures that all three stages are normalized in the same
way, removing from consideration differences that are only due to
normalization.
This feature is optional for now since it changes a low-level mechanism
and is not necessary for the majority of users. The "merge.renormalize"
config variable enables it.
Signed-off-by: Eyvind Bernhardsen <eyvind.bernhardsen@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* gv/portable:
test-lib: use DIFF definition from GIT-BUILD-OPTIONS
build: propagate $DIFF to scripts
Makefile: Tru64 portability fix
Makefile: HP-UX 10.20 portability fixes
Makefile: HPUX11 portability fixes
Makefile: SunOS 5.6 portability fix
inline declaration does not work on AIX
Allow disabling "inline"
Some platforms lack socklen_t type
Make NO_{INET_NTOP,INET_PTON} configured independently
Makefile: some platforms do not have hstrerror anywhere
git-compat-util.h: some platforms with mmap() lack MAP_FAILED definition
test_cmp: do not use "diff -u" on platforms that lack one
fixup: do not unconditionally disable "diff -u"
tests: use "test_cmp", not "diff", when verifying the result
Do not use "diff" found on PATH while building and installing
enums: omit trailing comma for portability
Makefile: -lpthread may still be necessary when libc has only pthread stubs
Rewrite dynamic structure initializations to runtime assignment
Makefile: pass CPPFLAGS through to fllow customization
Conflicts:
Makefile
wt-status.h
Introduce a new configuration variable, "core.eol", that allows the user
to set which line endings to use for end-of-line-normalized files in the
working directory. It defaults to "native", which means CRLF on Windows
and LF everywhere else.
Note that "core.autocrlf" overrides core.eol. This means that
[core]
autocrlf = true
puts CRLFs in the working directory even if core.eol is set to "lf".
Signed-off-by: Eyvind Bernhardsen <eyvind.bernhardsen@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Unfortunately, there are still plenty of production systems with
vendor compilers that choke unless all compound declarations can be
determined statically at compile time, for example hpux10.20 (I can
provide a comprehensive list of our supported platforms that exhibit
this problem if necessary).
This patch simply breaks apart any compound declarations with dynamic
initialisation expressions, and moves the initialisation until after
the last declaration in the same block, in all the places necessary to
have the offending compilers accept the code.
Signed-off-by: Gary V. Vaughan <gary@thewrittenword.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
As discussed on the list, "crlf" is not an optimal name. Linus
suggested "text", which is much better.
Signed-off-by: Eyvind Bernhardsen <eyvind.bernhardsen@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Change the semantics of the "crlf" attribute so that it enables
end-of-line normalization when it is set, regardless of "core.autocrlf".
Add a new setting for "crlf": "auto", which enables end-of-line
conversion but does not override the automatic text file detection.
Add a new attribute "eol" with possible values "crlf" and "lf". When
set, this attribute enables normalization and forces git to use CRLF or
LF line endings in the working directory, respectively.
The line ending style to be used for normalized text files in the
working directory is set using "core.autocrlf". When it is set to
"true", CRLFs are used in the working directory; when set to "input" or
"false", LFs are used.
Signed-off-by: Eyvind Bernhardsen <eyvind.bernhardsen@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Previously, autocrlf would only work well for normalized
repositories. Any text files that contained CRLF in the repository
would cause problems, and would be modified when handled with
core.autocrlf set.
Change autocrlf to not do any conversions to files that in the
repository already contain a CR. git with autocrlf set will never
create such a file, or change a LF only file to contain CRs, so the
(new) assumption is that if a file contains a CR, it is intentional,
and autocrlf should not change that.
The following sequence should now always be a NOP even with autocrlf
set (assuming a clean working directory):
git checkout <something>
touch *
git add -A . (will add nothing)
git commit (nothing to commit)
Previously this would break for any text file containing a CR.
Some of you may have been folowing Eyvind's excellent thread about
trying to make end-of-line translation in git a bit smoother.
I decided to attack the problem from a different angle: Is it possible
to make autocrlf behave non-destructively for all the previous problem cases?
Stealing the problem from Eyvind's initial mail (paraphrased and
summarized a bit):
1. Setting autocrlf globally is a pain since autocrlf does not work well
with CRLF in the repo
2. Setting it in individual repos is hard since you do it "too late"
(the clone will get it wrong)
3. If someone checks in a file with CRLF later, you get into problems again
4. If a repository once has contained CRLF, you can't tell autocrlf
at which commit everything is sane again
5. autocrlf does needless work if you know that all your users want
the same EOL style.
I belive that this patch makes autocrlf a safe (and good) default
setting for Windows, and this solves problems 1-4 (it solves 2 by being
set by default, which is early enough for clone).
I implemented it by looking for CR charactes in the index, and
aborting any conversion attempt if this is found.
Signed-off-by: Finn Arne Gangstad <finag@pvv.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If there are foreign $Id$ keywords in the repository, they are most
likely there for a reason. Let's keep them on checkout (which is also
what the documentation indicates). Foreign $Id$ keywords are now
recognized by there being multiple space separated fields in $Id:xxxxx$.
Signed-off-by: Henrik Grubbström <grubba@grubba.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
