Re: GCC 4 warnings in EPICS Base

[Andrew Johnson <anj@aps.anl.gov>]

Jeff Hill wrote:
> > I have reveived a bundle of patches to EPICS Base 3.14 that supposedly
> > fixes (most of) the GCC 4 warnings.
>
> I don't think that we have gcc 4.0 installed here yet. Is it being provided
> with the latest Linux? I just checked and it isn't provided yet by the
> cygwin installer.

It's shipped with the latest MacOS releases, and we have 4.0.2 installed 
here on our Solaris systems.

> > I have reveived a bundle of patches to EPICS Base 3.14 that supposedly
> > fixes (most of) the GCC 4 warnings.

I was sent a pile of warning patches a year or so ago which I looked at 
and didn't like some of them - any patches from non-core people need at 
least a review.  Anything that changes an API or public data structure 
should be posted here for wider comment first, and (unlikely here) for 
any new code that's longer than a couple of lines or so I'll need a 
Grant of License signed by the contributor's lab if I don't already have 
one.

One patch I remember definitely rejecting was a change to the gpHash 
structure that replaced the void* usrPvt pointer with a union of a 
function pointer and a void* pointer.  ANSI doesn't require these to be 
the same size so unfortunately something like this appears to be 
necessary, but the change would have broken any code outside of Base 
that uses gpHash.

> > Do you want me to ...
> > - commit them after quick review?
> > - commit them straight away?
> > - keep them for my own viewing pleasure?
> >
> > Should I
> > - create a special branch?
> > - tag this commit somehow specially?
> > - commit them to the 3.14 branch only?
> > - commit them to both 3.14 branch and 3.15 trunk?
> >
> > I'd like to hear your comments before completely messing things up
> > (which I'm going to do anyway, be assured).

I suggest review them yourself and commit anything straightforward to 
both the 3.14 branch and the trunk.  Anything unusual should be posted 
here for discussion.  I second Jeff's concerns.

> I am happy for you to commit a fix Ralph. It seems that it would be a waste
> to commit your fix to R3.14 w/o doing a merge of the R3.14 updates onto the
> main trunk also.
>
> Two concerns:
> 1) We need to be very careful not to add any new bugs to R3.14
> 2) There is always the peril that, when suppressing compiler warnings, we
> end up sweeping the dust bunnies under the carpet rather than eradicating
> them, and it might be better to have them in site (flagged by compiler
<pedant lang='en'>                             ^^^^ sight
</pedant>
> warnings) as opposed to having them lurk in dark places (suppressed by casts
> etc).

Thanks,

- Andrew
--
There is considerable overlap between the intelligence of the smartest
bears and the dumbest tourists -- Yosemite National Park Ranger