Problem reported by James Cook with reproduction scripts Regression test case by jamsek, as a simplified version of James' scripts fix tested + ok jamsek

"sed -i" is fundamentally unportable. GNU and OpenBSD sed(1) treat the extension for the backup file as an optional argument and use "sed -i" for no backup file. FreeBSD sed(1) treats the extension as an obligatory argument and uses "sed -i ''" for no backup file. There is no single syntax that works for both. ok stsp op

no need for sed (which doesn't change anything anyway), a no-op editor script is enough; the test ensures that the prepared log message is used. ok naddy@ stsp@

Ensure shell scripts can run under more POSIX-restricted shells. OK op@

ok stsp

This fixes the (harmless) errors raised by gitconfig.c on indented comment lines and adds a test case for it. Reported by James Cook, thanks! ok stsp@

conf_parse_line advances the `line' pointer without decrementing the line size `sz'. This makes the parsing code mistakingly reading from the next line (`line' is just a pointer in a bigger buffer that holds the whole file) and may mangle it by writing NUL bytes in it. Add also a new regress case to trigger this case. Reported by falsifian on IRC, thanks! ok stsp@

As per stsp's suggestion and building on his initial diff, add the -C option to enable creating commits with unresolved conflicts to the commit, histedit, merge, and rebase commands to allow continuing the operation despite files in conflict status. Also, only search for conflict markers in newly added lines to enable working with files already under version control that may have conflict markers embedded verbatim. lots of tweaks, improvements, and initial diff + ok stsp@

As discussed with op and stsp on irc, this prevents the valid case of changing a commit's author to yourself during a histedit operation, and potentially blocks other valid use cases too. ok stsp@ and op@

When run from the repository, display the UUID to help the user know which logmsg refs belong to which work tree. Also, use "backout" or "cherrypick" in the log message header to distinguish output from 'got log'. Suggested by stsp on irc. ok stsp@

This tests we correctly: - use multiple logmsg refs in one commit; - use only relevant refs involving affected paths; - use refs when some of the affected paths are reverted; - neither use nor remove refs if all affected paths are omitted from the commit; - use refs when previously excluded paths are later committed. - remove corresponding temp merged_log 'got-logmsg-*' files hints and ok stsp@

Allowing such author fields breaks 'got send' towards Github for affected commits because git-index-pack --strict will error out on the server: $ git index-pack --strict pack-de791fb6a3a1961e44ac5d98d72fd533bf9277c8.pack error: object 5d6bde9eaaf27f41ae8fa7112bb45e489d3c16b9: missingSpaceBeforeEmail: invalid author/committer line - missing space before email fatal: fsck error in packed object problem encountered by landry@ ok op@

The fix for CVE-2022-39253 in Git made our regression tests fail since creating submodules from local clones is no longer allowed by default. Add an override to our invocations of "git submodule add", the same which was added to Git's regression test suite as part of their fix. This CVE doesn't otherwise affect us. We do not implement submodules. Our use of them in regression tests exists only to ensure that Got does not freak out when it sees one. Precisely because automated support for nested repositories and/or working copies (as in Git Submodules, or Subversion Externals, etc.) have been an endless source of security problems in other systems. sudden test failures pointed out by naddy@

fixes test_rebase_no_author_info when a valid ~/.gitconfig is found. ok stsp@