Libgit2update

[Murmele]

• blame: fix update from buffer

  • branch: "origin/blame_buffer"
  • Another mr to upstream libgit2 with a similar patch: Git blame buffer gives the wrong result in many cases where there are… libgit2/libgit2#6572
  • Not needed anymore

• remote: add callbacks for connect/disconnect

  • pullrequest at libgit2 pending
  • branch: origin/callback_connect_disconnect

• hash: changed git_repository_hashfile to hash symlinks the same as when they are added to the index

  • branch: origin/hash

• submodule: pass absolute path to clone
  The init callback worked because it uses the parent repo path to contruct
  distinct repo and working directories, but the tests for an existing empty
  directory in git_clone were relative to the current working directory
  instead of the parent repository directory.

  • branch: origin/submodule
  • changed git_submodule_status to take a submodule instead of a repo and name
  • not needed anymore!

• patch: add context lines accessor

  • pullrequest at libgit2 pending
  • branch: 'origin/context_line_accessor´

• patch: add owner accessor

  • already upstream

• patch: added flag to disable mmap

  • really needed?

• push: fixed several update callback issues
  Update callback is no longer called (and the reference isn't updated) when the
  local and remote oids already match. The update callback is now called for tags.
  Finally, fixed a leak when there are multiple refspecs.

  • branch: fix_push_callback_issues

• diff: initialize git_diff_options.context_lines to an invalid default value and read 'diff.context' when the value is invalid

  • branch: "origin/diff_checkout"

• checkout: notify for deleted files

  • really needed?
  • branch: "origin/checkout_deletion_notification"

• blame: disable rename detection

  • really needed?
  • branch: "origin/disableRenameDetection"

• blame: add blame progress callback

  • submitted as pr to upstream
  • branch: origin/blame_abort

• fetch: work around failure to fetch annotated tags added to existing commits

  • branch: origin/fetch_annotated_tags

[Murmele]

@exactly-one-kas

2022-08-30T05:59:31.5311212Z C:\PROGRA~1\MICROS~2\2022\ENTERP~1\VC\Tools\Llvm\x64\bin\clang.exe -DWIN32 -D_WIN32_WINNT=0x0600 -ID:/a/Gittyup/Gittyup/build/release/dep/libgit2/libgit2/src/util -ID:/a/Gittyup/Gittyup/build/release/dep/libgit2/libgit2/include -ID:/a/Gittyup/Gittyup/build/release/dep/libgit2/libgit2/include/git2 -ID:/a/Gittyup/Gittyup/dep/libgit2/libgit2/src/libgit2 -ID:/a/Gittyup/Gittyup/dep/libgit2/libgit2/src/util -ID:/a/Gittyup/Gittyup/dep/libgit2/libgit2/include -D_GNU_SOURCE  -Wall -Wextra -Wdocumentation -Wno-documentation-deprecated-sync -Wno-missing-field-initializers -Wmissing-declarations -Wstrict-aliasing -Wstrict-prototypes -Wdeclaration-after-statement -Wshift-count-overflow -Wunused-const-variable -Wunused-function -Wint-conversion -Wc11-extensions -Wformat -Wformat-security -O3 -DNDEBUG -D_DLL -D_MT -Xclang --dependent-lib=msvcrt -Wall -Wextra -pedantic -Werror -std=c90 -MD -MT dep/libgit2/libgit2/tests/headertest/CMakeFiles/headertest.dir/headertest.c.obj -MF dep\libgit2\libgit2\tests\headertest\CMakeFiles\headertest.dir\headertest.c.obj.d -o dep/libgit2/libgit2/tests/headertest/CMakeFiles/headertest.dir/headertest.c.obj -c D:/a/Gittyup/Gittyup/dep/libgit2/libgit2/tests/headertest/headertest.c
2022-08-30T05:59:31.5352321Z In file included from D:/a/Gittyup/Gittyup/dep/libgit2/libgit2/tests/headertest/headertest.c:8:
2022-08-30T05:59:31.5365379Z In file included from D:/a/Gittyup/Gittyup/build/release/dep/libgit2/libgit2/include\git2.h:11:
2022-08-30T05:59:31.5368894Z In file included from D:/a/Gittyup/Gittyup/dep/libgit2/libgit2/include\git2/annotated_commit.h:11:
2022-08-30T05:59:31.5471330Z In file included from D:/a/Gittyup/Gittyup/dep/libgit2/libgit2/include\git2/repository.h:11:
2022-08-30T05:59:31.5474484Z D:/a/Gittyup/Gittyup/dep/libgit2/libgit2/include\git2/types.h:41:9: error: extension used [-Werror,-Wlanguage-extension-token]
2022-08-30T05:59:31.5475297Z typedef __int64 git_off_t;
2022-08-30T05:59:31.5475665Z         ^
2022-08-30T05:59:31.5476688Z D:/a/Gittyup/Gittyup/dep/libgit2/libgit2/include\git2/types.h:41:9: error: 'long long' is an extension when C99 mode is not enabled [-Werror,-Wlong-long]
2022-08-30T05:59:31.5477275Z 2 errors generated.

Do you understand why -Wlong-long -Werror is turned On on windows build?

[Murmele]

@exactly-one-kas are you using VS to build Gittyup? I'm using MinGW and there the Werror is not turned on

[exactly-one-kas]

MSVC works, and the error that required to use clang seems to be gone now, I'll switch back to MSVC

[Murmele]

I asked once in the libgit2 slack channel, but got no response

Hi, I am trying to move from git_filter.apply to git_filter.stream. I checked out the unittests to get a feeling how it should be. There I have seen for example in the wildcard test:

static int wildcard_filter_stream(
	git_writestream **out,
	git_filter *self,
	void **payload,
	const git_filter_source *src,
	git_writestream *next)
{
	return git_filter_buffered_stream_new(out,
		self, wildcard_filter_apply, NULL, payload, src, next);
}

git_filter_buffered_stream_new is used which is defined in filter.h. filter.h is not in the git2 folder so I thought it is not inteded to be used for the implementation? git_filter_buffered_stream_new needs a function pointer with git_str as arguments which are used normaly internally or?
What is the correct way to replace apply by filter streams?The problem is when I include filter.h in a C++ project it complains that in vector.h the keyword "new" is used as argument:

int git_vector_insert_sorted(git_vector *v, void *element,
	int (*on_dup)(void **old, void *new));
int git_vector_remove(git_vector *v, size_t idx);

[exactly-one-kas]

As I understand it, git_filter_buffered_stream_new is an internal libgit2 helper and we're supposed to malloc() a new git_writestream ourself (which will be free()'d by libgit2 after use)

Using the QProcess as the payload and hooking up the bytesWritten slot to write into next should work
git_filter.cleanup should then be a simple waiting to flush and delete

[Murmele]

self (which will be free()'d by libgit2 after use)

Thanks for the hint. I will check it out next days

[exactly-one-kas]

I don't think this is a good idea - this is an internal part of libgit2 and shouldn't be used by us

[exactly-one-kas]

Also, using git2 instead of libgit2 for target_link_libraries results in CMake trying to use the system-wide installed libgit2, so we should just put libgit2 in here

[exactly-one-kas]

I noticed there are very few differences between our fork and the upstream main branch
Do you think we can use the unmodified version and try to get the changes we need into the official libgit2?

[Murmele]

• Linux Flatpak
• Linux Appimage: SSH not detected
• Macos: SSH library not found
• Windows (Remove all files or choose a different installation folder and you see that the ssh lib is missing)