Re: [PATCH v5] [PR mi/29554] New PRINT-VALUES option '--scalar-values'.

[Gareth Rees <grees@undo.io>]

Eli Zaretskii wrote:
> The reason why I preferred #2 is simple: it avoids incompatible
> changes in the behavior of existing options.  Since it was not really
> clear-cut that the previous behavior was a bug, the backward
> incompatibility could cause trouble to some application or use case
> which didn't consider it was a bug and relied on that behavior.
> Introducing a new option is free from this problem.

Let me set out an argument for the existing behaviour being a bug. The
documentation for the -stack-list-arguments GDB/MI command [1] says:

> If print-values is [...] 2 or --simple-values, print the name, type
> and value for simple data types, and the name and type for arrays,
> structures and unions.

There is similar wording for the -stack-list-locals and
-stack-list-variables GDB/MI commands. The documentation does not say
what happens for reference types, but a natural way to read it is that
references to simple data types are simple, and references to compound
data types are not simple. That's certainly how I interpreted the
documentation when trying to fix the Visual Studio Code stack trace
performance issue [2]: my first attempt a pull request for this
problem used --simple-values [3] and you can see from the review
comments on the pull request that I wasn't the first developer to
interpret the documentation in this way: previously Visual Studio Code
had used --simple-values but this turned out not to work due to
references to compound types being considered simple [4].

This discrepancy between documentation and behaviour seems likely that
the current behaviour of --simple-values is a mistake (perhaps the
original implementer neglected to consider reference types, because
languages like C don't have them).

Anyway, this is not a blocker for me: I am happy to implement a new
option if that's what the GDB developers agree is the best approach.
But I also think the history of work on the Visual Studio Code issue
shows that the current behaviour of --simple-values is itself a bug
magnet.

[1] https://sourceware.org/gdb/onlinedocs/gdb/GDB_002fMI-Stack-Manipulation.html
[2] https://github.com/microsoft/MIEngine/issues/1349
[3] https://github.com/microsoft/MIEngine/pull/1350
[4] https://github.com/microsoft/MIEngine/pull/673