Re: [OSSTEST PATCH 6/9] db_retry: Suppress an "exiting via last" warning

From: Ian Campbell <ian.campbell@citrix.com>
To: Ian Jackson <Ian.Jackson@eu.citrix.com>
Cc: xen-devel@lists.xenproject.org
Subject: Re: [OSSTEST PATCH 6/9] db_retry: Suppress an "exiting via last" warning
Date: Thu, 17 Dec 2015 18:10:01 +0000	[thread overview]
Message-ID: <1450375801.1617.86.camel@citrix.com> (raw)
In-Reply-To: <22130.62853.68527.604758@mariner.uk.xensource.com>

On Thu, 2015-12-17 at 17:48 +0000, Ian Jackson wrote:
> Ian Campbell writes ("Re: [OSSTEST PATCH 6/9] db_retry: Suppress an
> "exiting via last" warning"):
> > On Thu, 2015-12-17 at 17:06 +0000, Ian Jackson wrote:
> > > This warning appears when db_retry_abort is used, since 2b069b6c
> > > "Database locking: Perl: Retry all deadlocks in PostgreSQL".
> > 
> > Is the reason for not turning this into a return related to the
> fact this
> > is within an eval, not a proper sub? But then I'm not really sure
> why it is
> > warning (I suppose eval has some sub like properties?).
> 
> I think it's a warning because eval does not trap `last's (nor
> `return's).  This means that if you think `eval' can be used to make
> try-finally, you will be disappointed.
> 
> (Unlike, in Tcl, say, where `break' and `continue' and indeed 
> `return
> are implemented as exceptions.)
> 
> Why would `return' be better ?  It would still generate the same
> warning.  Leaving the loop via `last' seems less heavyweight than
> using `return' and less likely to produce bugs in future patches whose
> authors don't spot the non-local exit.

I think I'm probably confused about the scope of the eval vs
{last,return} etc. In any case, that's no reason to block this patch
AFAICT: Acked-by: Ian Campbell <ian.campbell@citrix.com>

Ian.

> 
> > Related to 
> https://rt.perl.org/Public/Bug/Display.html?id=29238 perhaps.
> 
> I don't think so, although there seems to be a fair amount of
> confusion there.
> 
> Ian.