<html xmlns:o="urn:schemas-microsoft-com:office:office" xmlns:w="urn:schemas-microsoft-com:office:word" xmlns:m="http://schemas.microsoft.com/office/2004/12/omml" xmlns="http://www.w3.org/TR/REC-html40">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
<meta name="Generator" content="Microsoft Word 15 (filtered medium)">
<style><!--
/* Font Definitions */
@font-face
{font-family:"Cambria Math";
panose-1:2 4 5 3 5 4 6 3 2 4;}
@font-face
{font-family:Calibri;
panose-1:2 15 5 2 2 2 4 3 2 4;}
/* Style Definitions */
p.MsoNormal, li.MsoNormal, div.MsoNormal
{margin:0in;
margin-bottom:.0001pt;
font-size:12.0pt;
font-family:"Calibri",sans-serif;}
a:link, span.MsoHyperlink
{mso-style-priority:99;
color:#0563C1;
text-decoration:underline;}
a:visited, span.MsoHyperlinkFollowed
{mso-style-priority:99;
color:#954F72;
text-decoration:underline;}
span.EmailStyle17
{mso-style-type:personal-compose;
font-family:"Calibri",sans-serif;
color:windowtext;}
.MsoChpDefault
{mso-style-type:export-only;
font-family:"Calibri",sans-serif;}
@page WordSection1
{size:8.5in 11.0in;
margin:1.0in 1.0in 1.0in 1.0in;}
div.WordSection1
{page:WordSection1;}
--></style>
</head>
<body lang="EN-US" link="#0563C1" vlink="#954F72">
<div class="WordSection1">
<p class="MsoNormal"><span style="font-size:11.0pt">Hi all,<o:p></o:p></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt"><o:p> </o:p></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt">Currently, clang's format specifier check warnings (-Wformat) check for exact type matches between the specifier and the corresponding argument. I'm proposing adding an option to relax these warnings such
that it doesn't warn when there's a type mismatch but the types of the specifier and the argument have the same size and alignment (and are both integral). For an example, on an LLP64 system, attempting to print a long with "%d" would not warn, but attempting
to print a long long with "%d" would still warn. We could either do this relaxation for -Wformat and add a new option for the stricter warnings, or keep -Wformat as is and add a new option for the relaxed warnings.<o:p></o:p></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt"><o:p> </o:p></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt">The motivation is that during the 6.0 timeframe, clang gained format specifier checking for %zd (r308067) and %tu (r314470), which made -Wformat fire in a lot of additional cases. A particularly common case
is programs using %zd to print NSInteger. This will be warning-free on 64-bit targets, where both ssize_t and NSInteger are long, but warn on 32-bit targets, where ssize_t is still long but NSInteger is int. This additional warning motivated
<a href="https://reviews.llvm.org/D42933">https://reviews.llvm.org/D42933</a>, which attempted to special case %z and NSInteger; the diff's comments have a bunch of additional discussion about this issue. The conclusion there was that, although having mismatched
specifiers of this nature is definitely technically incorrect, there shouldn't actually be any issues in practice when the size and alignment of the specifier and the argument match. Ideally people would update their codebases to use proper specifiers, but
that can be impractical for large codebases, and it would be really nice to have a way to restrict the warning to the cases which will actually cause problems in practice.<o:p></o:p></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt"><o:p> </o:p></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt">Note that -Wformat isn't meant to be a general portability checker, since it only ever raises warnings when there's a type mismatch *on the current target*, not if there's any possibility of a mismatch. For
example, you could be printing out a size_t using %lu, and you wouldn't get warnings if your compilation target used unsigned long as the underlying type for size_t, even though that underlying type could differ across targets. This proposal does mean we would
lose out on some additional portability warnings; continuing the previous example, we would now not get warnings for printing out a size_t using %lu even for a compilation target where size_t was actually an unsigned int, assuming that the target had the same
size and alignment for int and long (as most 32-bit platforms do). However, given that -Wformat was never meant to be a general portability checker, and that there would still be an option to get the fully strict warnings, I don't think this is too much of
an issue in practice.<o:p></o:p></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt"><o:p> </o:p></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt">What are people's thoughts here? Are there preferences for relaxing -Wformat and adding a new option for strict format warnings vs. keeping -Wformat as-is and adding a new option for relaxed format warnings?
Of course, we should also ensure that the optimizer doesn't do anything surprising when there's a type mismatch between the specifier and the argument but both types have the same size and alignment (i.e., the case where the relaxed format warning won't fire),
both now and in the future.<o:p></o:p></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt"><o:p> </o:p></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt">Thanks,<o:p></o:p></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt">Shoaib<o:p></o:p></span></p>
</div>
</body>
</html>