[cfe-dev] Visit nodes in the order of the user output to create better reports

[Csaba Dabis via cfe-dev]

The current way of bug reporting:
---
1 Assuming 'i' is not equal to 1
----
2 Assuming 'i' is not equal to 2
3 Taking false branch
4 Taking false branch
---

If I would like to hook extra information in the linear backwards way, this
is going to be like the following:
---
1 Assuming 'i' is not equal to 1
----
2 Knowing 'i' is not equal to 2
3 Knowing 'i' is not equal to 1
4 Assuming 'i' is not equal to [1, 2]
---

And actually the proper approach would be:
---
1 Assuming 'i' is not equal to 1
---
2 Assuming 'i' is not equal to 2
3 Knowing 'i' is not equal to [1, 2]
4 Knowing 'i' is not equal to [1, 2]
---

George Karpenkov <ekarpenkov at apple.com> ezt írta (időpont: 2018. nov. 16.,
P, 19:04):

> Going forward is ambiguous, as there are many leafs, but only one root.
> What are you trying to achieve? Going backwards should never be a problem.
>
> > On Nov 16, 2018, at 8:21 AM, Csaba Dabis via cfe-dev <
> cfe-dev at lists.llvm.org> wrote:
> >
> > Hello!
> >
> > Why does the 'generateVisitorsDiagnostics()' function in BugReporter.cpp
> work backwards? It starts from the node of the error and iterate over the
> preds.
> >
> > I have tried to make it to work forward with storing the 'const
> ExplodedNode *' nodes in a vector, but 'auto P = V->VisitNode(NextNode,
> BRC, *R);' crashes. I just found out the 'LastPiece' checking could be
> moved out from the while-loop, that is it for now.
> >
> > Is it possible to achieve? What is could be problematic with that code
> snippet?
> >
> > Thanks you for the suggestions,
> > Csaba.
> > _______________________________________________
> > cfe-dev mailing list
> > cfe-dev at lists.llvm.org
> > http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20181116/b426837e/attachment.html>