From: Jeff King <peff@peff.net>
To: Johannes Schindelin <johannes.schindelin@gmx.de>
Cc: git@vger.kernel.org, Junio C Hamano <gitster@pobox.com>
Subject: Re: [PATCH v2 4/8] read_early_config(): optionally return the worktree's top-level directory
Date: Sat, 10 Jun 2017 05:44:24 -0400 [thread overview]
Message-ID: <20170610094423.kscoqrljxjki3np4@sigill.intra.peff.net> (raw)
In-Reply-To: <0cef47383c0dd2287ffda2a6eaf1f430ed3ee726.1496951503.git.johannes.schindelin@gmx.de>
On Thu, Jun 08, 2017 at 09:53:41PM +0200, Johannes Schindelin wrote:
> So far, when we invoked the early config code path, we implicitly
> determined the top-level directory of the worktree while discovering the
> .git/ directory.
>
> And then we simply forgot that information.
>
> However, when we expand aliases, we very much need that information, as
> aliases expanding to shell commands, i.e. whose value starts with an
> exclamation point, have to be executed in the top-level directory of the
> worktree. There are exceptions abound, not only with non-shell aliases
> (which are supposed to be executed in the original working directory
> instead), but also when being started inside the .git/ directory or in a
> worktree created via `git worktree add`.
I understand why you wrote it this way, but it really feels backwards.
We're calling a function that's about config, and oh by the way
sometimes it tells us about which git directory we're in (and sometimes
not because it didn't actually run discover_git_directory anyway).
It would make a lot more sense to me if we told read_early_config() "oh
by the way, I discovered the git directory already, here it is".
The same applies to the later patch to pass the information through
alias_lookup().
Taking a step back, this is really just an optimization, right? The
cleanest thing would be for the alias code, right before launching a
shell alias, to discover_git_directory(). But you don't want to do that
because it's too expensive?
If so, my questions would be:
1. Is it actually that expensive, or is this premature optimization?
We are, after all, about to fork and execute a shell. Have we
measured?
2. Can we cache the results of discovery_git_directory() between
calls? That would not only fix your issue here, but it would
optimize the calls we make when we call read_early_config() for
multiple items (e.g., both pager and alias config).
The only trick is making sure our previous value is still valid.
I suspect it would work to just ignore this, as we only chdir when
doing setup_git_directory(), and that should generally take
precedence over discover_git_directory() being called at all. But
for extra safety, we should be able to key it to the cwd, like:
strbuf_getcwd(&now_cwd);
if (!strbuf_cmp(&cached_cwd, &now_cwd) {
strbuf_addbuf(cdup_dir, &cached_cdup_dir));
return xstrdup(cached_gitdir);
}
strbuf_swap(&cached_cwd, &now_cwd);
strbuf_release(&now_cwd);
... rest of function, but store result in cached_gitdir ...
-Peff
next prev parent reply other threads:[~2017-06-10 9:44 UTC|newest]
Thread overview: 32+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-06-08 19:53 [PATCH v2 0/8] Avoid problem where git_dir is set after alias expansion Johannes Schindelin
2017-06-08 19:53 ` [PATCH v2 1/8] discover_git_directory(): avoid setting invalid git_dir Johannes Schindelin
2017-06-10 8:40 ` Jeff King
2017-06-08 19:53 ` [PATCH v2 2/8] config: report correct line number upon error Johannes Schindelin
2017-06-10 9:04 ` Jeff King
2017-06-13 11:02 ` Johannes Schindelin
2017-06-13 11:28 ` Jeff King
2017-06-08 19:53 ` [PATCH v2 3/8] help: use early config when autocorrecting aliases Johannes Schindelin
2017-06-10 9:05 ` Jeff King
2017-06-08 19:53 ` [PATCH v2 4/8] read_early_config(): optionally return the worktree's top-level directory Johannes Schindelin
2017-06-10 9:44 ` Jeff King [this message]
2017-06-08 19:53 ` [PATCH v2 5/8] t1308: relax the test verifying that empty alias values are disallowed Johannes Schindelin
2017-06-10 4:29 ` Junio C Hamano
2017-06-13 10:50 ` Johannes Schindelin
2017-06-13 12:59 ` Junio C Hamano
2017-06-08 19:53 ` [PATCH v2 6/8] t7006: demonstrate a problem with aliases in subdirectories Johannes Schindelin
2017-06-08 19:53 ` [PATCH v2 7/8] alias_lookup(): optionally return top-level directory Johannes Schindelin
2017-06-10 10:18 ` Jeff King
2017-06-13 11:21 ` Johannes Schindelin
2017-06-13 11:29 ` Jeff King
2017-06-13 11:42 ` Johannes Schindelin
2017-06-13 11:42 ` Jeff King
2017-06-13 16:40 ` Junio C Hamano
2017-06-14 6:02 ` Jeff King
2017-06-14 10:03 ` Johannes Schindelin
2017-06-08 19:53 ` [PATCH v2 8/8] Use the early config machinery to expand aliases Johannes Schindelin
2017-06-10 10:07 ` Jeff King
2017-06-10 10:10 ` Jeff King
2017-06-12 20:43 ` Junio C Hamano
2017-06-13 11:25 ` Johannes Schindelin
2017-06-13 11:30 ` Jeff King
2017-06-08 22:32 ` [PATCH v2 0/8] Avoid problem where git_dir is set after alias expansion Brandon Williams
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
List information: http://vger.kernel.org/majordomo-info.html
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20170610094423.kscoqrljxjki3np4@sigill.intra.peff.net \
--to=peff@peff.net \
--cc=git@vger.kernel.org \
--cc=gitster@pobox.com \
--cc=johannes.schindelin@gmx.de \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
Code repositories for project(s) associated with this public inbox
https://80x24.org/mirrors/git.git
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).