]> git.proxmox.com Git - libgit2.git/log
libgit2.git
9 years agolocal: use the packbuilder to push
Carlos Martín Nieto [Thu, 7 May 2015 11:44:28 +0000 (13:44 +0200)]
local: use the packbuilder to push

Instead of copying each object individually, as we'd been doing, use the
packbuilder which should be faster and give us some feedback.

While performing this change, we can hook up the packbuilder's writing
to the push progress so the caller knows how far along we are.

9 years agolocal: add clarification for non-bare push restriction
Carlos Martín Nieto [Thu, 7 May 2015 11:28:51 +0000 (13:28 +0200)]
local: add clarification for non-bare push restriction

9 years agoMerge pull request #3066 from libgit2/cmn/remote-less-state
Carlos Martín Nieto [Wed, 13 May 2015 08:47:13 +0000 (10:47 +0200)]
Merge pull request #3066 from libgit2/cmn/remote-less-state

Remove the configuration state we keep in the remote

9 years agopush: free the update list
Carlos Martín Nieto [Mon, 4 May 2015 07:39:55 +0000 (09:39 +0200)]
push: free the update list

9 years agoUpdate CHANGELOG with the changes to the remotes
Carlos Martín Nieto [Sun, 26 Apr 2015 00:20:22 +0000 (02:20 +0200)]
Update CHANGELOG with the changes to the remotes

9 years agoremote: simplify anonymous creation
Carlos Martín Nieto [Fri, 24 Apr 2015 22:38:22 +0000 (00:38 +0200)]
remote: simplify anonymous creation

We're down to simply having it be a call to create_internal() so let's
simply do that. The rest of the code is just a distraction.

9 years agoremote: remove git_remote_save()
Carlos Martín Nieto [Thu, 23 Apr 2015 04:55:29 +0000 (06:55 +0200)]
remote: remove git_remote_save()

It has now become a no-op, so remove the function and all references to
it.

9 years agoremote: remove live changing of refspecs
Carlos Martín Nieto [Thu, 23 Apr 2015 04:51:34 +0000 (06:51 +0200)]
remote: remove live changing of refspecs

The base refspecs changing can be a cause of confusion as to what is the
current base refspec set and complicate saving the remote's
configuration.

Change `git_remote_add_{fetch,push}()` to update the configuration
instead of an instance.

This finally makes `git_remote_save()` a no-op, it will be removed in a
later commit.

9 years agoexamples: adjust to the new remote API
Carlos Martín Nieto [Thu, 23 Apr 2015 04:01:13 +0000 (06:01 +0200)]
examples: adjust to the new remote API

9 years agoremote: move the tagopt setting to the fetch options
Carlos Martín Nieto [Wed, 22 Apr 2015 15:29:20 +0000 (17:29 +0200)]
remote: move the tagopt setting to the fetch options

This is another option which we should not be keeping in the remote, but
is specific to each particular operation.

9 years agoremote: move the update_fetchhead setting to the options
Carlos Martín Nieto [Wed, 22 Apr 2015 14:11:10 +0000 (16:11 +0200)]
remote: move the update_fetchhead setting to the options

While this will rarely be different from the default, having it in the
remote adds yet another setting it has to keep around and can affect its
behaviour. Move it to the options.

9 years agoremote: move the transport ctor to the callbacks
Carlos Martín Nieto [Wed, 22 Apr 2015 13:45:21 +0000 (15:45 +0200)]
remote: move the transport ctor to the callbacks

Instead of having it set in a different place from every other callback,
put it the main structure. This removes some state from the remote and
makes it behave more like clone, where the constructors are passed via
the options.

9 years agoremote: add prune option to fetch
Carlos Martín Nieto [Wed, 22 Apr 2015 02:54:00 +0000 (04:54 +0200)]
remote: add prune option to fetch

Add a prune setting in the fetch options to allow to fall back to the
configuration (the default) or to set it on or off.

9 years agoremote: remove url and pushurl from the save logic
Carlos Martín Nieto [Wed, 22 Apr 2015 02:38:08 +0000 (04:38 +0200)]
remote: remove url and pushurl from the save logic

As a first step in removing the repository-saving logic, don't allow
chaning the url or push url from a remote object, but change the
configuration on the configuration immediately.

9 years agoRemove the callbacks struct from the remote
Carlos Martín Nieto [Tue, 21 Apr 2015 20:10:36 +0000 (22:10 +0200)]
Remove the callbacks struct from the remote

Having the setting be different from calling its actions was not a great
idea and made for the sake of the wrong convenience.

Instead of that, accept either fetch options, push options or the
callbacks when dealing with the remote. The fetch options are currently
only the callbacks, but more options will be moved from setters and
getters on the remote to the options.

This does mean passing the same struct along the different functions but
the typical use-case will only call git_remote_fetch() or
git_remote_push() and so won't notice much difference.

9 years agopush: remove own copy of callbacks
Carlos Martín Nieto [Tue, 21 Apr 2015 18:16:48 +0000 (20:16 +0200)]
push: remove own copy of callbacks

The push object knows which remote it's associated with, and therefore
does not need to keep its own copy of the callbacks stored in the
remote.

Remove the copy and simply access the callbacks struct within the
remote.

9 years agoMerge pull request #3111 from whoisj/centralizing-buffer-sizes
Carlos Martín Nieto [Tue, 12 May 2015 09:27:37 +0000 (11:27 +0200)]
Merge pull request #3111 from whoisj/centralizing-buffer-sizes

Centralizing all IO buffer size values

9 years agoMerge pull request #3018 from ethomson/stash_apply
Carlos Martín Nieto [Tue, 12 May 2015 08:29:58 +0000 (10:29 +0200)]
Merge pull request #3018 from ethomson/stash_apply

Stash apply

9 years agocheckout: better document the `baseline_index` opt
Edward Thomson [Mon, 11 May 2015 18:10:24 +0000 (14:10 -0400)]
checkout: better document the `baseline_index` opt

9 years agostash application: document new API in CHANGELOG
Edward Thomson [Mon, 11 May 2015 18:06:47 +0000 (14:06 -0400)]
stash application: document new API in CHANGELOG

9 years agostash: propagate the error when writing a tree
Edward Thomson [Mon, 11 May 2015 18:02:53 +0000 (14:02 -0400)]
stash: propagate the error when writing a tree

9 years agostash_apply: provide progress callbacks
Edward Thomson [Fri, 1 May 2015 22:34:38 +0000 (18:34 -0400)]
stash_apply: provide progress callbacks

9 years agostash_apply: provide its own options structure
Edward Thomson [Fri, 1 May 2015 22:07:10 +0000 (18:07 -0400)]
stash_apply: provide its own options structure

9 years agostash: test checkout notify callbacks
Edward Thomson [Tue, 21 Apr 2015 00:06:02 +0000 (20:06 -0400)]
stash: test checkout notify callbacks

9 years agostash apply: default to at least GIT_CHECKOUT_SAFE
Edward Thomson [Tue, 21 Apr 2015 00:05:23 +0000 (20:05 -0400)]
stash apply: default to at least GIT_CHECKOUT_SAFE

9 years agostash: return GIT_EMERGECONFLICT on merge conflict
Edward Thomson [Wed, 18 Mar 2015 05:54:04 +0000 (01:54 -0400)]
stash: return GIT_EMERGECONFLICT on merge conflict

9 years agostash: document merge conflicts
Edward Thomson [Fri, 1 May 2015 17:53:46 +0000 (13:53 -0400)]
stash: document merge conflicts

9 years agostash: refactor to use merge_iterators
Edward Thomson [Thu, 5 Mar 2015 04:55:42 +0000 (23:55 -0500)]
stash: refactor to use merge_iterators

9 years agostash: ensure a reflog has entries
Edward Thomson [Wed, 18 Feb 2015 21:33:50 +0000 (16:33 -0500)]
stash: ensure a reflog has entries

9 years agostash apply: check out a tree, not piecewise
Edward Thomson [Wed, 18 Feb 2015 20:16:05 +0000 (15:16 -0500)]
stash apply: check out a tree, not piecewise

9 years agostash: use git_commit_summary for a summary
Edward Thomson [Wed, 18 Feb 2015 18:24:07 +0000 (13:24 -0500)]
stash: use git_commit_summary for a summary

9 years agoindex: introduce git_index_read_index
Edward Thomson [Wed, 18 Mar 2015 04:25:18 +0000 (00:25 -0400)]
index: introduce git_index_read_index

9 years agocheckout: allow baseline to be specified as index
Edward Thomson [Mon, 16 Mar 2015 22:57:57 +0000 (18:57 -0400)]
checkout: allow baseline to be specified as index

Allow the baseline to be specified as an index, so that users
need not write their index to a tree just to checkout with that
as the baseline.

9 years agomerge: merge iterators
Edward Thomson [Thu, 19 Feb 2015 03:53:40 +0000 (22:53 -0500)]
merge: merge iterators

9 years agoAdded git_stash_apply() and git_stash_pop() APIs
Pierre-Olivier Latour [Fri, 14 Nov 2014 03:32:47 +0000 (12:32 +0900)]
Added git_stash_apply() and git_stash_pop() APIs

9 years agocentralizing all IO buffer size values
J Wyman [Mon, 11 May 2015 17:19:25 +0000 (10:19 -0700)]
centralizing all IO buffer size values

9 years agoMerge pull request #3104 from whoisj/optimal-buffer-size
Carlos Martín Nieto [Sat, 9 May 2015 11:11:46 +0000 (13:11 +0200)]
Merge pull request #3104 from whoisj/optimal-buffer-size

Adjusting stream buffer size to 64KB

9 years agoAdjusting stream buffer size to 64KB
J Wyman [Thu, 7 May 2015 15:50:12 +0000 (08:50 -0700)]
Adjusting stream buffer size to 64KB

64K is optimal buffer size per https://technet.microsoft.com/en-us/library/cc938632.aspx

9 years agoMerge pull request #3086 from yongthecoder/master
Carlos Martín Nieto [Wed, 6 May 2015 16:39:39 +0000 (18:39 +0200)]
Merge pull request #3086 from yongthecoder/master

Android build doesn't need deps/regex

9 years agoMerge pull request #3100 from arthurschreiber/patch-5
Carlos Martín Nieto [Wed, 6 May 2015 05:37:49 +0000 (07:37 +0200)]
Merge pull request #3100 from arthurschreiber/patch-5

Ensure frameworks are mentioned in libgit2.pc

9 years agoEnsure frameworks are mentioned in libgit2.pc
Arthur Schreiber [Tue, 5 May 2015 20:36:24 +0000 (22:36 +0200)]
Ensure frameworks are mentioned in libgit2.pc

When building on Mac OS X, the `CoreFoundation` and `Security` frameworks where missing from `Libs.private` in the generated `libgit2.pc` file.

9 years agoconfig: plug a couple of leaks
Carlos Martín Nieto [Tue, 5 May 2015 07:47:16 +0000 (09:47 +0200)]
config: plug a couple of leaks

9 years agoAndroid build doesn't need deps/regex
Yong Li [Mon, 4 May 2015 13:41:34 +0000 (09:41 -0400)]
Android build doesn't need deps/regex

deps/regex was included in Android build because Android NDK 4 has
a packaging bug and doesn't have the regular expression functions defined
in its libc.so. The bug has been fixed in subsequent Android NDK releases.

If it is still necessary to work around the bug in Android NDK 4, we
should consider to use an option like ANDROID_NDK_RELEASE or
ANDROID_NDK_RELEASE_NUM.

9 years agoMerge pull request #3079 from ethomson/config
Carlos Martín Nieto [Mon, 4 May 2015 13:23:40 +0000 (15:23 +0200)]
Merge pull request #3079 from ethomson/config

Configuration changes for handling multiple of the same sections

9 years agoMerge pull request #3070 from ethomson/checkout_icase
Carlos Martín Nieto [Mon, 4 May 2015 12:59:08 +0000 (14:59 +0200)]
Merge pull request #3070 from ethomson/checkout_icase

Case insensitive checkout improvements

9 years agoconfig: test all multivars are updated
Edward Thomson [Mon, 4 May 2015 12:36:21 +0000 (07:36 -0500)]
config: test all multivars are updated

If a multivar exists within two sections (of the same name)
then they should both be updated in a `set_multivar`.  Ensure
that this is the case.

9 years agoconfig: use wildcard in test instead of empty expr
Edward Thomson [Mon, 27 Apr 2015 20:31:18 +0000 (16:31 -0400)]
config: use wildcard in test instead of empty expr

9 years agoconfig: cleanup some now-unused variables
Edward Thomson [Mon, 27 Apr 2015 20:29:00 +0000 (16:29 -0400)]
config: cleanup some now-unused variables

9 years agoconfig: lock the file for write before reading
Edward Thomson [Mon, 27 Apr 2015 19:38:44 +0000 (15:38 -0400)]
config: lock the file for write before reading

When writing a configuration file, we want to take a lock on the
new file (eg, `config.lock`) before opening the configuration file
(`config`) for reading so that we can prevent somebody from changing
the contents underneath us.

9 years agoTest setting config var under duplicate header.
Ryan Roden-Corrent [Tue, 21 Apr 2015 19:57:20 +0000 (15:57 -0400)]
Test setting config var under duplicate header.

Add a test that exposes a bug in config_write.
It is valid to have multiple separate headers for the same config section, but
config_write will exit after finding the first matching section in certain
situations.
This test proves that config_write will duplicate a variable that already
exists instead of overwriting it if the variable is defined under a duplicate
section header.

9 years agoconfig: write existing lines as-is when rewriting
Edward Thomson [Mon, 27 Apr 2015 14:43:50 +0000 (10:43 -0400)]
config: write existing lines as-is when rewriting

When updating a configuration file, we want to copy the old data
from the file to preserve comments and funny whitespace, instead
of writing it in some "canonical" format.  Thus, we keep a
pointer to the start of the line and the line length to preserve
these things we don't care to rewrite.

9 years agoconfig: examine whole file when writing
Edward Thomson [Thu, 23 Apr 2015 20:54:36 +0000 (16:54 -0400)]
config: examine whole file when writing

Previously we would try to be clever when writing the configuration
file and try to stop parsing (and simply copy the rest of the old
file) when we either found the value we were trying to write,
or when we left the section that value was in, the assumption being
that there was no more work to do.

Regrettably, you can have another section with the same name later
in the file, and we must cope with that gracefully, thus we read the
whole file in order to write a new file.

Now, writing a file looks even more than reading.  Pull the config
parsing out into its own function that can be used by both reading
and writing the configuration.

9 years agoconfig: test overwriting cvar in multiple regions
Edward Thomson [Thu, 23 Apr 2015 16:00:05 +0000 (12:00 -0400)]
config: test overwriting cvar in multiple regions

9 years agoconfig: ensure we can write to an empty file
Edward Thomson [Tue, 21 Apr 2015 21:18:21 +0000 (17:18 -0400)]
config: ensure we can write to an empty file

9 years agocheckout test: better case-insensitive test on Mac
Edward Thomson [Thu, 23 Apr 2015 19:58:53 +0000 (15:58 -0400)]
checkout test: better case-insensitive test on Mac

On Mac OS, `realpath` is deficient in determining the actual filename
on-disk as it will simply provide the string you gave it if that file
exists, instead of returning the filename as it exists.  Instead we
must read the directory entries for the parent directory to get the
canonical filename.

9 years agostatus test: always test the new file path
Edward Thomson [Tue, 31 Mar 2015 20:06:33 +0000 (16:06 -0400)]
status test: always test the new file path

9 years agocheckout test: only run icase on icase platform
Edward Thomson [Thu, 23 Apr 2015 13:21:33 +0000 (09:21 -0400)]
checkout test: only run icase on icase platform

9 years agocheckout: break case-changes into delete/add
Edward Thomson [Tue, 31 Mar 2015 20:29:35 +0000 (16:29 -0400)]
checkout: break case-changes into delete/add

When checking out with a case-insensitive working directory, we
want to change the case of items in the working directory to
reflect changes that occured in the checkout target.  Diff now
has an option to break case-changing renames into delete/add.

9 years agocheckout: remove blocking dir when FORCEd
Edward Thomson [Tue, 31 Mar 2015 20:28:13 +0000 (16:28 -0400)]
checkout: remove blocking dir when FORCEd

9 years agoRevert "Always checkout with case sensitive iterator"
Edward Thomson [Fri, 27 Mar 2015 19:39:28 +0000 (15:39 -0400)]
Revert "Always checkout with case sensitive iterator"

This reverts commit 40d791545abfb3cb71553a27dc64129e1a9bec28.

9 years agocheckout test: ensure we write to casechanged dir
Edward Thomson [Wed, 1 Apr 2015 19:23:37 +0000 (15:23 -0400)]
checkout test: ensure we write to casechanged dir

Ensure that on a case insensitive filesystem that we can checkout
into some folder 'FOLDER' that exists on disk, even if the target
of the checkout is a different case (eg 'folder').

9 years agocheckout test: ignore unstaged case-changing renames
Edward Thomson [Fri, 27 Mar 2015 19:33:44 +0000 (15:33 -0400)]
checkout test: ignore unstaged case-changing renames

On Windows, you might sloppily rewrite a file (or have a sloppy
text editor that does it for you) and accidentally change its
case.  (eg, "README" -> "readme").  Git ignores this accidental
case changing rename during checkout and will happily write the
new content to the file despite the name change.  We should, too.

9 years agoMerge pull request #3087 from ethomson/pr/3054
Carlos Martín Nieto [Mon, 4 May 2015 09:16:52 +0000 (11:16 +0200)]
Merge pull request #3087 from ethomson/pr/3054

Performance Improvements to Status on Windows

9 years agoMerge pull request #3089 from volftomas/patch-2
Edward Thomson [Sat, 2 May 2015 14:51:24 +0000 (10:51 -0400)]
Merge pull request #3089 from volftomas/patch-2

Added call to git_libgit2_shutdown()

9 years agoRestructured to be nicer example
Tomas Paladin Volf [Sat, 2 May 2015 10:16:22 +0000 (12:16 +0200)]
Restructured to be nicer example

Code restructured to better represent best practice when
using libgit2.

9 years agoMerge pull request #3092 from techee/iconv_macports_fix2
Edward Thomson [Fri, 1 May 2015 21:30:51 +0000 (17:30 -0400)]
Merge pull request #3092 from techee/iconv_macports_fix2

Don't search iconv in /opt/local

9 years agoDon't search iconv in /opt/local
Jiří Techet [Fri, 1 May 2015 20:48:33 +0000 (22:48 +0200)]
Don't search iconv in /opt/local

Since OpenSSL isn't used any more on OS X, there is no dependency
on any MacPorts library under /opt/local and there is no danger of
conflicts between MacPorts and system iconv. For this reason the
system iconv can always be used now.

9 years agodiriter: actually use iconv on mac
Edward Thomson [Fri, 1 May 2015 15:27:44 +0000 (11:27 -0400)]
diriter: actually use iconv on mac

9 years agogit_buf_put_w: introduce utf16->utf8 conversion
Edward Thomson [Wed, 29 Apr 2015 22:12:51 +0000 (18:12 -0400)]
git_buf_put_w: introduce utf16->utf8 conversion

9 years agogit_path_diriter: use FindFirstFile in win32
Edward Thomson [Wed, 29 Apr 2015 21:23:02 +0000 (17:23 -0400)]
git_path_diriter: use FindFirstFile in win32

Using FindFirstFile and FindNextFile in win32 allows us to
use the directory information that is returned, instead of
us having to get the file attributes all over again, which
is a distinct cost savings on win32.

9 years agogit_path_diriter: next shouldn't take path ptr
Edward Thomson [Wed, 29 Apr 2015 18:31:59 +0000 (14:31 -0400)]
git_path_diriter: next shouldn't take path ptr

The _next method shouldn't take a path pointer (and a path_len
pointer) as 100% of current users use the full path and ignore
the filename.

Plus let's add some docs and a unit test.

9 years agogit_path_dirload_with_stat: moved to fs_iterator
Edward Thomson [Wed, 29 Apr 2015 18:04:01 +0000 (14:04 -0400)]
git_path_dirload_with_stat: moved to fs_iterator

9 years agogit_path_dirload_with_stat: use git_path_diriter
Edward Thomson [Wed, 29 Apr 2015 17:28:29 +0000 (13:28 -0400)]
git_path_dirload_with_stat: use git_path_diriter

9 years agogit_win32_path_dirload_with_stat: removed
Edward Thomson [Wed, 29 Apr 2015 18:03:20 +0000 (14:03 -0400)]
git_win32_path_dirload_with_stat: removed

9 years agogit_path_dirload: use git_path_diriter
Edward Thomson [Wed, 29 Apr 2015 15:58:10 +0000 (11:58 -0400)]
git_path_dirload: use git_path_diriter

9 years agogit_path: introduce 'git_path_diriter'
Edward Thomson [Wed, 29 Apr 2015 15:05:27 +0000 (11:05 -0400)]
git_path: introduce 'git_path_diriter'

Introduce a new `git_path_diriter` that can iterate directories
efficiently for each platform.

9 years agoMerge pull request #3059 from libgit2/cmn/negotiation-notify
Edward Thomson [Fri, 1 May 2015 15:28:54 +0000 (11:28 -0400)]
Merge pull request #3059 from libgit2/cmn/negotiation-notify

[WIP/RFC] push: report the update plan to the caller

9 years agoMerge pull request #3081 from leoyanggit/build_warnings
Edward Thomson [Fri, 1 May 2015 13:01:34 +0000 (09:01 -0400)]
Merge pull request #3081 from leoyanggit/build_warnings

Fix some build warnings

9 years agoMerge pull request #3090 from libgit2/jamill/doc_update
Carlos Martín Nieto [Thu, 30 Apr 2015 19:55:10 +0000 (21:55 +0200)]
Merge pull request #3090 from libgit2/jamill/doc_update

Update documentation for API changes

9 years agoMerge pull request #3088 from volftomas/patch-1
Carlos Martín Nieto [Thu, 30 Apr 2015 19:25:54 +0000 (21:25 +0200)]
Merge pull request #3088 from volftomas/patch-1

Fixed unused warning in tests/rebase/merge.c

9 years agoUpdate documentation for API changes
Jameson Miller [Thu, 30 Apr 2015 14:57:13 +0000 (10:57 -0400)]
Update documentation for API changes

9 years agoAdded call to git_libgit2_shutdown()
Tomas Paladin Volf [Thu, 30 Apr 2015 13:10:28 +0000 (15:10 +0200)]
Added call to git_libgit2_shutdown()

Added forgotten call to git_libgit2_shutdown() to the /examples/network/git2.c.

9 years agoFixed unused warning in tests/rebase/merge.c
Tomas Paladin Volf [Thu, 30 Apr 2015 12:17:05 +0000 (14:17 +0200)]
Fixed unused warning in tests/rebase/merge.c

9 years agoMerge pull request #3085 from yongthecoder/regfree
Carlos Martín Nieto [Wed, 29 Apr 2015 20:08:46 +0000 (22:08 +0200)]
Merge pull request #3085 from yongthecoder/regfree

Do not call regfree() on an empty regex that is not successfully created...

9 years agoDo not call regfree() on an empty regex that is not successfully created by regcomp
Yong Li [Wed, 29 Apr 2015 17:40:42 +0000 (13:40 -0400)]
Do not call regfree() on an empty regex that is not successfully created by regcomp

(also removed an unused member "has_regex" from all_iter)

9 years agowin32: keep full path for realpath usage
Edward Thomson [Tue, 28 Apr 2015 20:39:47 +0000 (16:39 -0400)]
win32: keep full path for realpath usage

9 years agowin32: mimic git_path_dirload_with_stat closely
Edward Thomson [Tue, 28 Apr 2015 16:24:08 +0000 (12:24 -0400)]
win32: mimic git_path_dirload_with_stat closely

9 years agodirload: loop conditional; less path mangling
Edward Thomson [Tue, 28 Apr 2015 15:16:42 +0000 (11:16 -0400)]
dirload: loop conditional; less path mangling

9 years agowin32 dirload: don't heap allocate DIR structure
Edward Thomson [Mon, 27 Apr 2015 22:02:06 +0000 (18:02 -0400)]
win32 dirload: don't heap allocate DIR structure

9 years agowin32: abstract file attributes -> struct stat fn
Edward Thomson [Mon, 27 Apr 2015 21:47:51 +0000 (17:47 -0400)]
win32: abstract file attributes -> struct stat fn

9 years agoImprovements to status performance on Windows.
J Wyman [Thu, 26 Mar 2015 22:10:24 +0000 (18:10 -0400)]
Improvements to status performance on Windows.

Changed win32/path_w32.c to utilize NTFS' FindFirst..FindNext data instead of doing an lstat per file. Avoiding unnecessary directory opens and file scans reduces IO, improving overall performance. Effect is magnified due to NTFS being a kernel mode file system (as opposed to user mode).

9 years agoImprovements to ignore performance on Windows.
J Wyman [Mon, 30 Mar 2015 21:07:44 +0000 (14:07 -0700)]
Improvements to ignore performance on Windows.

Minimizing the number directory and file opens, minimizes the amount of IO thus reducing the overall cost of performing ignore operations.

9 years agoFix some build warnings
Leo Yang [Tue, 28 Apr 2015 16:40:20 +0000 (12:40 -0400)]
Fix some build warnings

In checkout.c and filter.c we were casting a sub struct
to a parent struct which breaks the strict aliasing rules
in C. However we can use .parent or .base to access the
parent struct to avoid the build warnings.

In remote.c the local variable error was not initialized
or updated in some cases. For unintialized error a build
warning will be generated. So always keep error variable
up-to-date.

9 years agoMerge pull request #3071 from linquize/git_reflog_drop
Edward Thomson [Fri, 24 Apr 2015 16:00:51 +0000 (12:00 -0400)]
Merge pull request #3071 from linquize/git_reflog_drop

Fix wrong format string in git_reflog_drop() error message

9 years agoFix wrong format string in git_reflog_drop() error message
Linquize [Fri, 24 Apr 2015 15:56:59 +0000 (23:56 +0800)]
Fix wrong format string in git_reflog_drop() error message

9 years agoMerge pull request #3063 from ethomson/config_validate_name
Carlos Martín Nieto [Fri, 24 Apr 2015 00:46:49 +0000 (02:46 +0200)]
Merge pull request #3063 from ethomson/config_validate_name

Validate configuration keys

9 years agoMerge pull request #2997 from libgit2/cmn/secure-transport
Edward Thomson [Thu, 23 Apr 2015 19:24:05 +0000 (15:24 -0400)]
Merge pull request #2997 from libgit2/cmn/secure-transport

Use SecureTransport on OS X

9 years agoSecureTransport: handle graceful closes
Carlos Martín Nieto [Mon, 13 Apr 2015 13:39:58 +0000 (15:39 +0200)]
SecureTransport: handle graceful closes

On close, we might get a return code which looks like an error but just
means that the other side closed gracefully. Handle that.

9 years agoSecureTransport: require TLS v1.x
Carlos Martín Nieto [Tue, 24 Mar 2015 15:31:51 +0000 (16:31 +0100)]
SecureTransport: require TLS v1.x

Anything SSL is deprecated. Let's make sure we don't try to use SSL v3
when talking to the server.

9 years agoUpdate THREADING and CHANGELOG with SecureTransport details
Carlos Martín Nieto [Thu, 19 Mar 2015 13:26:07 +0000 (14:26 +0100)]
Update THREADING and CHANGELOG with SecureTransport details