[PATCH] Do not trust PWD blindly

[PATCH] Do not trust PWD blindly

[Johannes Schindelin]

At least on Windows, chdir() does not update PWD. Unfortunately, stat()
does not fill any ino or dev fields anymore, so get_pwd_cwd() is not
able to tell.

But there is a telltale: both ino and dev are 0 when they are not filled
correctly, so let's be extra cautious.

This happens to fix a bug in "get-receive-pack working_directory/" when
the GIT_DIR would not be set correctly due to absolute_path(".")
returning the wrong value.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
---

	On Fri, 8 Jul 2011, Pat Thoyts wrote:

	> ! t5516-fetch-push      (60 receive.denyCurrentBranch = updateInstead)

	This patch fixes that.

	Hannes, I have no idea whether you meant 10c4c881 to fix anything 
	on Windows.

 abspath.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)

diff --git a/abspath.c b/abspath.c
index 01858eb..37287f8 100644
--- a/abspath.c
+++ b/abspath.c
@@ -102,7 +102,8 @@ static const char *get_pwd_cwd(void)
 	pwd = getenv("PWD");
 	if (pwd && strcmp(pwd, cwd)) {
 		stat(cwd, &cwd_stat);
-		if (!stat(pwd, &pwd_stat) &&
+		if ((cwd_stat.st_dev || cwd_stat.st_ino) &&
+		    !stat(pwd, &pwd_stat) &&
 		    pwd_stat.st_dev == cwd_stat.st_dev &&
 		    pwd_stat.st_ino == cwd_stat.st_ino) {
 			strlcpy(cwd, pwd, PATH_MAX);
-- 
1.7.6.rc0.4047.g15f89

Re: [PATCH] Do not trust PWD blindly

[Sebastian Schuberth]

On Sat, Jul 9, 2011 at 19:38, Johannes Schindelin
<Johannes.Schindelin@gmx.de> wrote:

>        On Fri, 8 Jul 2011, Pat Thoyts wrote:
>
>        > ! t5516-fetch-push      (60 receive.denyCurrentBranch = updateInstead)
>
>        This patch fixes that.

I can confirm that the patch fixes test 5516 on Windows. Thanks Dscho!

-- 
Sebastian Schuberth

Re: [PATCH] Do not trust PWD blindly

[Johannes Sixt]

Am 09.07.2011 19:38, schrieb Johannes Schindelin:
> 
> At least on Windows, chdir() does not update PWD.

Very strange wording. chdir() should not update PWD even on POSIX.

> Unfortunately, stat()
> does not fill any ino or dev fields anymore, so get_pwd_cwd() is not
> able to tell.
> 
> But there is a telltale: both ino and dev are 0 when they are not filled
> correctly, so let's be extra cautious.
> 
> This happens to fix a bug in "get-receive-pack working_directory/" when
> the GIT_DIR would not be set correctly due to absolute_path(".")
> returning the wrong value.
> 
> Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
> ---
> 
> 	On Fri, 8 Jul 2011, Pat Thoyts wrote:
> 
> 	> ! t5516-fetch-push      (60 receive.denyCurrentBranch = updateInstead)
> 
> 	This patch fixes that.
> 
> 	Hannes, I have no idea whether you meant 10c4c881 to fix anything 
> 	on Windows.

I think this fix worked for me because when git is called from CMD, PWD
is not in the enviornment and the if (pwd && ...) branch is never taken.

> 
>  abspath.c |    3 ++-
>  1 files changed, 2 insertions(+), 1 deletions(-)
> 
> diff --git a/abspath.c b/abspath.c
> index 01858eb..37287f8 100644
> --- a/abspath.c
> +++ b/abspath.c
> @@ -102,7 +102,8 @@ static const char *get_pwd_cwd(void)
>  	pwd = getenv("PWD");
>  	if (pwd && strcmp(pwd, cwd)) {
>  		stat(cwd, &cwd_stat);
> -		if (!stat(pwd, &pwd_stat) &&
> +		if ((cwd_stat.st_dev || cwd_stat.st_ino) &&
> +		    !stat(pwd, &pwd_stat) &&
>  		    pwd_stat.st_dev == cwd_stat.st_dev &&
>  		    pwd_stat.st_ino == cwd_stat.st_ino) {
>  			strlcpy(cwd, pwd, PATH_MAX);

Acked-by: Johannes Sixt <j6t@kdbg.org>

-- Hannes

Re: [PATCH] Do not trust PWD blindly

[Johannes Schindelin]

Hi,

On Sun, 10 Jul 2011, Johannes Sixt wrote:

> Am 09.07.2011 19:38, schrieb Johannes Schindelin:
> > 
> > At least on Windows, chdir() does not update PWD.
> 
> Very strange wording. chdir() should not update PWD even on POSIX.

Well, you might think it is strange wording. But then, it expresses 
exactly what I meant it to say. chdir() does not update PWD on Windows.

You might be very surprised, but that is not true on the Linux system 
where one of the 4msysgit.git test cases does _not_ break, while it does 
on Windows.

I hoped to make that clear with the wording, but apparently I failed 
rather blatantly.

All the more surprising do I find that:

> Acked-by: Johannes Sixt <j6t@kdbg.org>

Ciao,
Johannes

Re: [PATCH] Do not trust PWD blindly

[Randal L. Schwartz]

>>>>> "Johannes" == Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:

Johannes> You might be very surprised, but that is not true on the Linux system 
Johannes> where one of the 4msysgit.git test cases does _not_ break, while it does 
Johannes> on Windows.

If you ever depend on a userspace PWD to be your actual current
directory without at least stat()ing it, you've failed.

In my experience, it is *never* reliable.  It's just a hint.

-- 
Randal L. Schwartz - Stonehenge Consulting Services, Inc. - +1 503 777 0095
<merlyn@stonehenge.com> <URL:http://www.stonehenge.com/merlyn/>
Smalltalk/Perl/Unix consulting, Technical writing, Comedy, etc. etc.
See http://methodsandmessages.posterous.com/ for Smalltalk discussion

Re: [PATCH] Do not trust PWD blindly

[Johannes Schindelin]

Hi Randal,

On Sun, 10 Jul 2011, Randal L. Schwartz wrote:

> >>>>> "Johannes" == Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:
> 
> Johannes> You might be very surprised, but that is not true on the Linux 
> Johannes> system where one of the 4msysgit.git test cases does _not_ 
> Johannes> break, while it does on Windows.
> 
> If you ever depend on a userspace PWD to be your actual current 
> directory without at least stat()ing it, you've failed.
> 
> In my experience, it is *never* reliable.  It's just a hint.

To be precise, get_pwd_cwd() _does_ stat() what's in PWD, and _does_ 
compare with the stat() of what comes out of getcwd(), but that comparison 
uses only st_dev and st_ino, both of which happen to be 0 in my case -- 
for each and every file/directory.

I can only _guess_ at the reasoning behind get_pwd_cwd(). I _think_ it was 
meant to catch the case when getcwd() and PWD refer to the same directory, 
but PWD goes through symbolic links. I was tempted to just throw that PWD 
handling out for Windows, since we do not have symbolic link handling yet. 
But that is currently actively discussed, so we might need it in the 
future, in which case I have to figure out how to fake reliable st_dev and 
st_ino values into our stat() code.

Ciao,
Dscho

Re: [PATCH] Do not trust PWD blindly

[Junio C Hamano]

Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:

> On Sun, 10 Jul 2011, Randal L. Schwartz wrote:
> ...
>> If you ever depend on a userspace PWD to be your actual current 
>> directory without at least stat()ing it, you've failed.
>> 
>> In my experience, it is *never* reliable.  It's just a hint.
>
> To be precise, get_pwd_cwd() _does_ stat() what's in PWD, and _does_ 
> compare with the stat() of what comes out of getcwd(), but that comparison 
> uses only st_dev and st_ino, both of which happen to be 0 in my case -- 
> for each and every file/directory.
>
> I can only _guess_ at the reasoning behind get_pwd_cwd(). I _think_ it was 
> meant to catch the case when getcwd() and PWD refer to the same directory, 
> but PWD goes through symbolic links.

Thanks for a much clearer explanation than before. I tried to reword the
proposed commit log message using the description above.

I feel that the title is still not optimal. If the original code used to
return getenv("PWD") if the environment variable is set, and otherwise
fell back to getcwd(), and the updated code tries to make sure they refer
to the same directory, then "Do not trust PWD blindly" would be a good
description for the fix, but the code you fixed the bug in tried not to
trust PWD blindly but failed to realize that on some systems dev/ino field
may be unreliable.

"Do not trust st.st_ino/st.st_dev blindly" might be a better title in that
sense.

In any case, thanks for a fix; will queue.

Author: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Date:   Sat Jul 9 19:38:08 2011 +0200

    Do not trust PWD blindly
    
    10c4c88 (Allow add_path() to add non-existent directories to the path,
    2008-07-21) introduced get_pwd_cwd() function in order to favor $PWD when
    getenv("PWD") and getcwd() refer to the same directory but are different
    strings (e.g. the former gives a nicer looking name via a symbolic link to
    an uglier looking automounted path). The function tried to determine if
    two directories are the same by running stat(2) on both and comparing
    ino/dev fields.
    
    Unfortunately, stat() does not fill any ino or dev fields in msysgit.  But
    there is a telltale: both ino and dev are 0 when they are not filled
    correctly, so let's be extra cautious.
    
    This happens to fix a bug in "get-receive-pack working_directory/" when
    the GIT_DIR would not be set correctly due to absolute_path(".")
    returning the wrong value.
    
    Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
    Acked-by: Johannes Sixt <j6t@kdbg.org>
    Signed-off-by: Junio C Hamano <gitster@pobox.com>

Re: [PATCH] Do not trust PWD blindly

[Johannes Schindelin]

Hi,

On Mon, 11 Jul 2011, Junio C Hamano wrote:

> "Do not trust st.st_ino/st.st_dev blindly" might be a better title in 
> that sense.

Maybe prefix it with "get_pwd_cwd(): "?

> In any case, thanks for a fix; will queue.

Thanks!
Johannes