Re: Re*: [PATCH] change contract between system_path and it's callers

[Junio C Hamano <gitster@pobox.com>]

Alexander Kuleshov <kuleshovmail@gmail.com> writes:

> Ah, you little ahead me, so we do not care about config.c in this way,
> because it takes git_etc_gitconfig() in the same way as
> git_etc_gitattributes

Yes, but looking at the file, you would notice that the
"use_local_config" codepath assigns a string obtained from
git_pathdup() to given_config_source.file, which means that somebody
later in the code path would not know if given_config_source.file is
merely pointing at a piece of memory owned by somebody else
(e.g. came from git_etc_gitconfig()) or if it owns the piece of
memory (e.g. came from git_pathdup()).  In the former case the
memory should never be freed, but not freeing in the latter would
leak the memory.

Assignment to given_config_source.file I see in that function that
borrows (i.e. the current code that does not fee is correct) are:

  - getenv(CONFIG_ENVIRONMENT)
  - NULL ;-) obviously
  - git_etc_gitconfig()
  
All the other assignment to given_config_source.file (including the
ones obtained from a call to home_config_paths()) makes us
responsible to free that piece memory, so they are technically
leaks.

But cmd_config() is a moral equivalent of main() in a typical
program, so it might not be worth worrying about allocating a single
piece of memory that is used throughout the lifetime of that
function and leaving it without freeing.