C++ Logo

sg15

Advanced search

Re: [Tooling] Ideas on debugging functions that have been 100% removed from the EXE

From: Mathias Stearn <redbeard0531_at_[hidden]>
Date: Sat, 16 Feb 2019 16:30:06 -1000
I know this isn't exactly what you are talking about, but gdb has a related
feature called xmethods:
https://sourceware.org/gdb/onlinedocs/gdb/Xmethods-In-Python.html. And
dwarf when emitted at a high enough level lets you single step linewise
through inlined functions.

Neither of these really help with constexpr though.

On Sat, Feb 16, 2019, 2:13 PM Scott Wardle <swardle_at_[hidden] wrote:

> Hi everyone,
>
> A few people in the SG14 group suggested I cross post this idea here as
> this is a debugging tools problem not just a Games/Trading/ low latency
> issue.
>
> Anyways I realized the other I have never seen a ISO paper that included
> the work debug in the title. This surprised me. Maybe they just don’t
> really go in the standard as it is more a quality issue than a wording
> one.
>
> My dream is to come up with a system where we can debug an optimized build
> like a debug build? Some mixed compile time vs runtime debugger debugger
> system. So even if the function is gone maybe there is some way to still
> step into it and see what the compiler saw. With a JIT compiler I am
> surprised this does not happen today. If you step in to a java function you
> might as well debug the debug one or have a choice between the debug and
> optimized ones. In C++ I think this is a harder problem… but I can’t tell
> if it is imposable.
>
> This is kind of a high bar maybe we could just do some of the easier
> function first? If we really want normally programmers to use meta classes,
> write JSON reader and REGEX-ing at compile time… I would say sooner or
> later it might be a good idea to step though the code in a debugger.
>
> Anyways here is my blog post.
> http://www.swardle.com/sweb/blog5.html
>
> Scott
> _______________________________________________
> Tooling mailing list
> Tooling_at_[hidden]
> http://www.open-std.org/mailman/listinfo/tooling
>

Received on 2019-02-17 03:30:20