On 10/31/2013 08:22 PM, Craig A. Berry wrote: > On Thu, Oct 31, 2013 at 11:27 AM, Karl Williamson > <public@khwilliamson.com> wrote: >> On 10/30/2013 10:36 PM, Father Chrysostomos via RT wrote: >>> >>> On Wed Oct 30 19:09:06 2013, craig.a.berry@gmail.com wrote: >>>> >>>> On Wed, Oct 30, 2013 at 3:31 PM, Father Chrysostomos via RT >>>> <perlbug-followup@perl.org> wrote: >>>>> >>>>> On Wed Oct 30 13:27:30 2013, sprout wrote: >>>> >>>> >>>>>> Revised patch attached. >>>>> >>>>> >>>>> No, that was the same patch. Revised patch *now* attached. >>>> >>>> >>>> Looks good to me. >>>> >>> >>> Now in as f789f6a4bdb. >>> >> >> I don't think we should be recommending -Accflags=-Werror=conversion >> until blead can be compiled using it, starting with Configure. Here's the >> output of make -k (after Configure) > > Sigh. I should've known the tiny test program I used that beautifully > displayed the shortening problem didn't represent the real world. > Even -Wconversion (without escalating the warning to an error) > includes a huge pile of different, unrelated potential problems when > thrown at the Perl sources. The closest we could come might be: > > -Accflags='-Wconversion -Wno-sign-conversion -Wno-shorten-64-to-32' > > which is kind of a mouthful and still generates so many warnings with > existing code that it would be hard to see any new ones, even if > people could develop the habit of reviewing and acting on warnings, > which experience shows usually won't happen. > > So I agree with you, but I don't know what else to recommend. Maybe there are real bugs here, so that we should look at. > > Side note: -Wshorten-64-to-32 might be useful when chasing the I32 bug > on a platform with -Duse64bitint enabled. >Thread Previous | Thread Next