Re: [PATCH] tools/toollog: Drop XTL_NEW_LOGGER()

From: Andrew Cooper <andrew.cooper3@citrix.com>
To: Ian Campbell <ian.campbell@citrix.com>,
	Xen-devel <xen-devel@lists.xen.org>
Cc: Wei Liu <wei.liu2@citrix.com>, Ian Jackson <Ian.Jackson@eu.citrix.com>
Subject: Re: [PATCH] tools/toollog: Drop XTL_NEW_LOGGER()
Date: Tue, 19 Jan 2016 16:40:24 +0000	[thread overview]
Message-ID: <569E66F8.2020407@citrix.com> (raw)
In-Reply-To: <1453220650.29930.107.camel@citrix.com>

On 19/01/16 16:24, Ian Campbell wrote:
> On Thu, 2016-01-14 at 20:13 +0000, Andrew Cooper wrote:
>> XTL_NEW_LOGGER() makes a number of unreasonable assumptions about the symbols
>> visible in its scope,
> It assumes that the function names to fill in the vtable and the type name
> are related, that hardly seems totally "unreasonable". What else does it
> assume that makes it unreasonable?

We have already had this argument once, the result being "patch
welcome".  Here one is.

Not only does it assume certain names, it is tokenised with a magic 3rd
identifier.

It also assumes the presence of a structure member named vtable.

>
> I think if you intend to remove something on this basis you need to be
> specific about what you believe the short comings are.

GNUism in header file claiming C99 strictness

If vtable isn't the first element in the structure, it follows a wild
pointer on error.

>
>>  and as such is only usable by its sole caller.
> "not usable by every imaginable caller" is not the same as "usable by one
> single possible caller", I think you are overstating the case here.

It is genuinely harder to reverse engineer how to use that macro than to
opencode a sane alternative.

I stand firmly by my statement.

~Andrew