[OE-core] Toolchain library whitelisting: A first pass (preliminary patch/RFC)

[Peter Seebach]

On Thu, 26 Apr 2012 17:01:41 -0500
Mark Hatle <mark.hatle at windriver.com> wrote:

> split does a whitespace based split automatically, I'm used to
> seeing .split() everywhere.  (I won't comment on the other similar
> split items)

Okay.
 
> > +    validities = data.getVarFlags('TUNEVALID') or ""
 
> "validities"?  that a new word?  ;)

Pretty much.  I was trying to think of a name for "the set of valid
things".  :)

> Change the above to:
>      multilibs = data.getVar("MULTILIBS", True)
>      if multilibs:

If someone has done:

MULTILIBS = ""

this then ends up being confused because pairs[0] of the single
returned item is still "", and that's not valid.
 
> > +		if pairs[1] == 'lib':
> > +		    tune_error_set.append("The multilib 'lib' was
> > specified, but that doesn't work. You need lib32 or lib64.")
 
> I'm surprised, why doesn't 'lib' work?  I was under the impression
> the naming was completely arbitrary.

I am surprised too, but if I use that name, I get a cryptic parse
error from libxcb's recipe which I couldn't understand.
 
> Also we can definitely have more then just lib32 or lib64.  We
> already often use libx32 in testing the x32 layer.

Okay.  I can improve the message, for sure.
 
> If the tune isn't defined, then the multilib configuration is
> invalid.  I thought we already had a check for that somewhere else..
> but if not.. it wouldn't be a bad idea to mention that here for the
> user.

That probably wants to be tested too.

> I wonder if this is an error or a warning.. I suspect it would be
> unintentional, but I'm not sure it's a failure?

I'm not totally sure.  I blamed that for my problem where I kept
getting a TUNE_ARCH of "x86_64x86_64" or "i586i586", but it turned out
to be a red herring.

I could change that to a warning.

> Your whitespace usage looks different.. perhaps that is just my
> mailer.

I almost certainly have tabs.

-s
-- 
Listen, get this.  Nobody with a good compiler needs to be justified.