[Lldb-commits] [PATCH] D66654: Prevent D66398 `TestDataFormatterStdList`-like regressions in the future

Jan Kratochvil via Phabricator via lldb-commits lldb-commits at lists.llvm.org
Mon Aug 26 13:01:25 PDT 2019

jankratochvil added a comment.

In D66654#1645658 <https://reviews.llvm.org/D66654#1645658>, @clayborg wrote:

> What about having the first one that matched being the one that is used and avoid errors all together?

This is how it worked so far before the fix D66398 <https://reviews.llvm.org/D66398> (which made the regexes unambiguous - always matching at most 1 regex for any string).  The previous version of the patch <https://reviews.llvm.org/D66398?id=215895> just sorted the regexes alphabetically because original LLDB did not sort them at all printing random results.

@labath did not like the alphabetical sorting as (IIUC) adding formatters from a new library can affect formatting for a different library.

> We do this with the regex alias stuff already.

I haven't found the work `alias` in either `grep -ri alias $(find -name DataFormatters)` or in  Variable Formatting doc <https://lldb.llvm.org/use/variable.html>.

> With regular expressions people are sure to make ones that conflict with each other.

So which order to choose? It should not have different output during different runs of LLDB.




More information about the lldb-commits mailing list