[LLVMdev] [lldb-dev] Crash reporting in clang/lldb

Zachary Turner zturner at google.com
Fri May 22 15:16:40 PDT 2015


On Mon, May 18, 2015 at 6:59 PM <jingham at apple.com> wrote:

>
> > On May 18, 2015, at 4:31 PM, Vince Harron <vince at nethacker.com> wrote:
> >
> > +llvmdev
> > -llvm-dev
> >
> >
> > On Mon, May 18, 2015 at 4:28 PM, Vince Harron <vince at nethacker.com>
> wrote:
> > Hi all,
> >
> > I would like to integrate crash reporting into Android developer tools,
> starting with lldb and clang.  Google Breakpad seems like a reasonable
> option.
> >
> > https://code.google.com/p/google-breakpad/
> >
> > Of course, it will only be enabled when the user has allowed it.
> >
> > The LLDB project is interested but it makes sense to us to push crash
> reporting down into llvm to make it available to all llvm projects.
> >
> > There are a couple of options here:
> >
> > 1) integrate this only downstream in our repo
> > 2) integrate in lldb ToT, defaulted to off by #defines
> > 3) integrate in lldb ToT without #define requirements
> >
> > I would prefer #3
> >
> > Thoughts?
> >
> >
>
> I would prefer #3.  You already have to assemble a bunch of pieces to
> start doing lldb development, I'd rather not add another.
>
> Also, I'd prefer to have this OFF for debug builds: getting reports of
> development lldb's crashing is not helpful for anything other than telling
> whether I (or you, or Greg...) are having a good day...
>
> If the "allow crash reporting" is a system or user-wide setting, then the
> best way to do this and NOT force all lldb developers to turn off
> crash-reporting for the lldb installed on their system is to have some
> defines that are not set in the debug build.
>

What about a debugger setting like debugger.enable-crash-reporting which
defaults to False in debug builds and true in release builds.  Or just
always defaults to False unless you build with a specific preprocessor
define, but you can still turn it on in the application by changing the
setting.

Plus this gives users a convenient means to opt out of it, which is usually
desirable when it comes to things that collect data.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20150522/6640c46b/attachment.html>


More information about the llvm-dev mailing list