[cfe-dev] [analyzer] Checker identifiers

[Sean Eveson via cfe-dev]

>
> Hi! A patch about bug identification just got accepted (
> http://reviews.llvm.org/D10305), however one of the weakness of the
> current
> method is the lack of unique identifiers for static analyzer checkers, that
> remain unchanged over time. Right now the checkers are identified by full
> name including packages, which is not a robust solution, because these
> checkers are sometimes moved to a separate package. I would like to
> propose a new requirement for the names of the checkers. If
> the checker names (without the package names) would be unique and unchanged
> over time, it would be possible to use the names in the issue hashes and
> moving a checker to another package would not affect the hashes. If you
> find this requirement reasonable, I will propose a patch which makes
> checker name collision a hard error (even when they are in different
> package). In case you think this requirement is overly restrictive an
> alternative way
> would be to have an extra string for each checker which can be similar to
> git commit hashes. What do you think? Thanks,
> Gabor


Hi Gabor.


> I would like to propose a new requirement for the names of the checkers.
> If the checker names (without the package names) would be unique and
> unchanged over time, it would be possible to use the names in the issue
> hashes and moving a checker to another package would not affect the hashes.


Would checker names (without the package names) end up containing redundant
information to avoid collisions? e.g. Would Unix.API have to become
something like Unix.UnixAPI to avoid a collision with OSX.API?

In case you think this requirement is overly restrictive an alternative way
> would be to have an extra string for each checker which can be similar to
> git commit hashes.
>

This sounds better to me, or maybe a GUID?

Thanks,

Sean Eveson
SN Systems - Sony Computer Entertainment
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20151016/216eaa4c/attachment.html>