Re: [PATCH v4 2/2] gitk: synchronize config write

From: Paul Mackerras <paulus@samba.org>
Date: Mar 2, 2015, 00:10 UTC
Message-ID: <20150302001050.GC24862@iris.ozlabs.ibm.com>
In-Reply-To: <1415571602-5858-3-git-send-email-max@max630.net>

On Mon, Nov 10, 2014 at 12:20:02AM +0200, Max Kirillov wrote:

Show 19 quoted lines

> If several gitk instances are closed simultaneously, safestuff procedure
> can run at the same time, resulting in a conflict which may cause losing
> of some of the instance's changes, failing the saving operation or even
> corrupting the configuration file. This can happen, for example, at user
> session closing, or at group closing of all instances of an application
> which is possible in some desktop environments.
> 
> To avoid this, make sure that only one saving operation is in progress.
> It is guarded by existance of $config_file_tmp file. Both creating the
> file and moving it to $config_file are atomic operations, so it should
> be reliable.
> 
> Reading does not need to be syncronized, because moving is atomic
> operation, and the $config_file always refers to full and correct file.
> But, if there is a stale $config_file_tmp file, report it at gitk start.
> If such file is detected at saving, just abort the saving, because this
> is how gitk used to handle errors while saving.
> 
> Signed-off-by: Max Kirillov <max@max630.net>

The idea looks good; I have a couple of comments on the patch. First, 50 tries over 5 seconds seems a bit excessive to me, wouldn't (say) 20 tries be enough? Is the 50 the result of some analysis?

> +	    error_popup "Probably there is stale $config_file_tmp file; config saving is going to fail. Check if it is being used by any existing gitk process and remove it otherwise"

I would word this as "There appears to be a stale $config_file_tmp file, which will prevent gitk from saving its configuration on exit. Please remove it if it is not being used by any existing gitk process."

Show 29 quoted lines

> @@ -2811,11 +2824,16 @@ proc savestuff {w} {
>  
>      if {$stuffsaved} return
>      if {![winfo viewable .]} return
> +    set remove_tmp 0
>      catch {
> -	if {[file exists $config_file_tmp]} {
> -	    file delete -force $config_file_tmp
> +	set try_count 0
> +	while {[catch {set f [open $config_file_tmp {WRONLY CREAT EXCL}]}]} {
> +	    if {[incr try_count] > 50} {
> +		error "Unable to write config file: $config_file_tmp exists"
> +	    }
> +	    after 100
>  	}
> -	set f [open $config_file_tmp w]
> +	set remove_tmp 1
>  	if {$::tcl_platform(platform) eq {windows}} {
>  	    file attributes $config_file_tmp -hidden true
>  	}
> @@ -2878,6 +2896,14 @@ proc savestuff {w} {
>  	puts $f "}"
>  	close $f
>  	file rename -force $config_file_tmp $config_file
> +	set remove_tmp 0
> +	return ""
> +    } err
> +    if {$err ne ""} {
> +	puts "Error saving config: $err"

I would suggest checking the return from the catch statement, like this:

	if {[catch {
	    ...
	    file rename -force $config_file_tmp $config_file
	} err]} {
	    puts "Error saving config: $err"
	    if {$remove_tmp} {
		file delete -force $config_file_tmp
	    }
	}

rather than doing a return inside the catch.

Paul.