Gé Weijers, Thank you for your detailed explanation.
My understanding of this question is at a different level with your
generous help.
I agree with what all you said indeed.
I still have a question, how C/C++ achieve this goal (i.e. reporting
multiple errors after running a compilation)?
Lua and all the "checking code" that has ever been written probably comes in at 10 MB (an incoherent guess obviously). The LLVM source code is over 1.2 GB.
Thanks to all of you.
Best regards
sunshilong
On Wed, Jan 27, 2021 at 2:39 AM Gé Weijers <ge@weijers.org> wrote:
>
> On Sun, Jan 24, 2021 at 11:20 PM 孙世龙 sunshilong <sunshilong369@gmail.com> wrote:
> >
> > Hi, list
> > As the title, It seems impossible to achieve this goal since Lua
> > stops to do more work after encountering the first error.
> > Am I missing something?
>
> The problem is that there is no good and efficient strategy to modify
> the input ('fix' the error) so the compiler is guaranteed to be able
> to continue and analyze the rest of the program in a meaningful way.
> What do you do, insert a symbol, delete a symbol, ...? If the 'fix' is
> wrong you're likely to end up with a cascade of error messages that
> are completely missing the mark.
>
> In the days when compiling a program took a long time there was a good
> reason to pay attention to recovering from compilation errors so you
> could fix multiple errors between compiler runs, but compiling a Lua
> program takes milliseconds on any reasonable machine so it's probably
> not worth the trouble to modify the parser to do something clever
> enough to recover from most errors.
>
> --
> Gé
