From: Junio C Hamano <gitster@pobox.com>
Date: Mon, 21 Jul 2014 17:04:54 GMT
Subject: Re: [PATCH v2 2/2] Make locked paths absolute when current directory is changed
Message-ID: <xmqqr41ek5hl.fsf@gitster.dls.corp.google.com>
In-Reply-To: <CACsJy8AXc4jvLPNpGyGdY9uzrnN-SbEeiksLDpS_=29gJ1KMnQ@mail.gmail.com>

Duy Nguyen <pclouds@gmail.com> writes:

> On Mon, Jul 21, 2014 at 8:27 PM, Ramsay Jones
> <ramsay@ramsay1.demon.co.uk> wrote:
>>> +void make_locked_paths_absolute(void)
>>> +{
>>> +     struct lock_file *lk;
>>> +     for (lk = lock_file_list; lk != NULL; lk = lk->next) {
>>> +             if (lk->filename && !is_absolute_path(lk->filename)) {
>>> +                     char *to_free = lk->filename;
>>> +                     lk->filename = xstrdup(absolute_path(lk->filename));
>>> +                     free(to_free);
>>> +             }
>>> +     }
>>> +}
>>
>> I just have to ask, why are we putting relative pathnames in this
>> list to begin with? Why not use an absolute path when taking the
>> lock in all cases? (calling absolute_path() and using the result
>> to take the lock, storing it in the lock_file list, should not be
>> in the critical path, right? Not that I have measured it, of course! :)
>
> Conservative :) I'm still scared from 044bbbc (Make git_dir a path
> relative to work_tree in setup_work_tree() - 2008-06-19). But yeah
> looking through "grep hold_" I think none of the locks is in critical
> path. absolute_path() can die() if cwd is longer than PATH_MAX (and
> doing this reduces the chances of that happening). But René is adding
> strbuf_getcwd() that can remove that PATH_MAX. So I guess we should be
> fine with putting absolute_path() in hold_lock_file_...*

OK, we should center these efforts around the strbuf_getcwd() topic,
basing the other topic on realpath() and this one on it then?