[Lldb-commits] [lldb] r327356 - [ExpressionParser] Fix crash when evaluating invalid expresssions.
Davide Italiano via lldb-commits
lldb-commits at lists.llvm.org
Tue Mar 13 10:37:24 PDT 2018
An alternative proposed by Fred which is an OK middle ground IMHO is
that of not inserting a decl for the unmangled name, but treat this
symbol always as-it-is.
i.e. if we have a symbol _Znwm in some object, we don't insert a decl
for `operator new(unsigned long)` but for _Znwm itself.
Greg, Jason, would that work for you guys?
On Tue, Mar 13, 2018 at 10:25 AM, Davide Italiano <dccitaliano at gmail.com> wrote:
> On Tue, Mar 13, 2018 at 9:59 AM, Frédéric Riss via lldb-commits
> <lldb-commits at lists.llvm.org> wrote:
>>> On Mar 12, 2018, at 6:40 PM, Davide Italiano via lldb-commits <lldb-commits at lists.llvm.org> wrote:
>>> Author: davide
>>> Date: Mon Mar 12 18:40:00 2018
>>> New Revision: 327356
>>> URL: http://llvm.org/viewvc/llvm-project?rev=327356&view=rev
>>> [ExpressionParser] Fix crash when evaluating invalid expresssions.
>>> Typical example, illformed comparisons (operator== where LHS and
>>> RHS are not compatible). If a symbol matched `operator==` in any
>>> of the object files lldb inserted a generic function declaration
>>> in the ASTContext on which Sema operates. Maintaining the AST
>>> context invariants is fairly tricky and sometimes resulted in
>>> crashes inside clang (or assertions hit).
>>> The real reason why this feature exists in the first place is
>>> that of allowing users to do something like:
>>> (lldb) call printf("patatino")
>>> even if the debug informations for printf() is not available.
>>> Eventually, we might reconsider this feature in its
>>> entirety, but for now we can't remove it as it would break
>>> a bunch of users. Instead, try to limit it to non-C++ symbols,
>>> where getting the invariants right is hopefully easier.
>>> Now you can't do in lldb anymore
>>> (lldb) call _Zsomethingsomething(1,2,3)
>>> but that doesn't seem to be such a big loss.
>> I’m somewhat surprised by this. My understanding of the crash you were investigating is that Clang crashed because we injected a Decl looking like this: “void operator==(…)” after finding the operator== symbol somewhere. I think injecting bogus C++ Decls makes no sense and it cannot really work anyway.
>> On the other hand, I see no harm in injecting “_Zsomethingsomething(…)” as a C Decl. This can be useful, and people should be able to call raw symbols in their binaries. Is there no way to keep the later while preventing the creation of broken C++ decls?
> Thank you all for your feedback. I'll reply with a single mail to everybody.
> C decls can be inserted. In fact, this works, even after my changes:
> (lldb) call printf("patatino")
> (int) $0 = 8
> I always thought identifiers beginning with underscore where illegal in C.
> Here's what the standard says:
> Section 7.1.3
> "All identifiers that begin with an underscore and either an uppercase
> letter or another underscore are always reserved for any use."
> They're not quite illegal, but they're reserved, so I'm unsure how
> frequent having symbols starting with `_Z` is popular.
> Maybe lldb has a better way of detecting whether this is a C or a C++ program?
> There are several constraints:
> 1) The object from which we're loading symbols has no debug info, so
> we can't look at the CU and just say whether it's C or C++ or
> 2) The expression parser always evaluates expressions in a C++ context
> (to the best of my understanding)
> 3) You can always mix-and-match C/C++ object files as they're just
> Mach-O or ELF objects at that point (not recommended, but I've seen
> people doing it).
> Do you have any thoughts on how this should be achieved?
More information about the lldb-commits