[PATCH] [StaticAnalyzer] New checker Sizeof on expression

Anders Rönnholm Anders.Ronnholm at evidente.se
Tue May 13 03:27:27 PDT 2014

Från: Anders Rönnholm
Skickat: den 27 mars 2014 11:09
Till: Jordan Rose
Cc: cfe-commits at cs.uiuc.edu; Daniel Marjamäki
Ämne: SV: [PATCH] [StaticAnalyzer] New checker Sizeof on expression

New patch with new diagnostic message. I couldn't come up with a better wording so i'm using your suggestion. I don't know of a good way to silence the warning.

I removed the check for HasSideEffects previously but had to take back.  I noticed that the patch triggered some false positives without it.


Anders Rönnholm Senior Engineer
Evidente ES East AB  Warfvinges väg 34  SE-112 51 Stockholm  Sweden

Mobile:                    +46 (0)70 912 42 54
E-mail:                    Anders.Ronnholm at evidente.se


Från: Jordan Rose [jordan_rose at apple.com]
Skickat: den 31 januari 2014 18:50
Till: Anders Rönnholm
Cc: cfe-commits at cs.uiuc.edu; Daniel Marjamäki
Ämne: Re: [PATCH] [StaticAnalyzer] New checker Sizeof on expression

Sorry to have let this slip! This is looking good, but I had one more thought about the diagnostic message. It says "may yield unexpected results", but doesn't really explain what those unexpected results are. I was wondering if we could work the type into the message for the operator case.

"operand of sizeof is a binary expression of type %0, which may not be intended"

I don't like that wording either, but at least this one makes people say "what? why isn't it [the type I actually want]?". Also, should there be a way to silence the warning?

What do you think?

On Jan 23, 2014, at 6:40 , Anders Rönnholm <Anders.Ronnholm at evidente.se<mailto:Anders.Ronnholm at evidente.se>> wrote:


New one with comments handled.

Från: Jordan Rose [jordan_rose at apple.com<mailto:jordan_rose at apple.com>]
Skickat: den 20 december 2013 19:15
Till: Anders Rönnholm
Cc: cfe-commits at cs.uiuc.edu<mailto:cfe-commits at cs.uiuc.edu>; Daniel Marjamäki; Anna Zaks; David Blaikie; Richard Smith; Matt Calabrese
Ämne: Re: [PATCH] [StaticAnalyzer] New checker Sizeof on expression

On Dec 10, 2013, at 4:38 , Anders Rönnholm <Anders.Ronnholm at evidente.se<mailto:Anders.Ronnholm at evidente.se><mailto:Anders.Ronnholm at evidente.se>> wrote:

Are you OK to commit this patch or do you see more issues?

I'm not sure if anyone else has ideological concerns. There's always a flag to turn this off, I suppose.

+  if (S.isSFINAEContext())
+      return;

Code style: extra indent?

+  if(E->HasSideEffects(S.getASTContext()))
+    return;

sizeof doesn't evaluate its argument, so I'm not sure why you wouldn't want to warn here.

+  const FunctionDecl *FD = S.getCurFunctionDecl();
+  if(FD && FD->isFunctionTemplateSpecialization())
+    return;

Code style: space after if. (Above too, actually.)

+    if (Binop->getLHS()->getType()->isArrayType() ||
+        Binop->getLHS()->getType()->isAnyPointerType() ||
+        Binop->getRHS()->getType()->isArrayType() ||
+        Binop->getRHS()->getType()->isAnyPointerType())
+      return;

I don't think this is correct...the user is only trying to get ptrdiff_t if both the LHS and RHS are pointer-ish.

+def warn_sizeof_bin_op : Warning<
+  "using sizeof() on an expression with an operator may yield unexpected results">,
+  InGroup<SizeofOnExpression>;
+def warn_sizeof_sizeof : Warning<
+  "using sizeof() on sizeof() may yield unexpected results.">,
+  InGroup<SizeofOnExpression>;

sizeof doesn't actually require parens, so we shouldn't put the parens in the diagnostics.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: sizeofonexpression.diff
Type: text/x-patch
Size: 8147 bytes
Desc: sizeofonexpression.diff
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20140513/dc257547/attachment.bin>

More information about the cfe-commits mailing list