Entry points are currently named by the "canonical name" listed in the pPrettyName field. These names are intended for forward-compatibility cases. They are designed to be human-readable but only within the context of an entry point, having them "loose" like this is confusing.
Entry points are shown in the bottom-up and flat views. Even more confusingly they end up merged into a single row with 100% cost, with an arbitrary name in the scope column.
To Reproduce
Occurs in all multithreaded meta.db databases.
Expected behavior
Entry points should be named with names befitting their context:
When possible a proper name should be generated based on the entryPoint field (e.g. (entryPoint: 1) maps to <program root>).
If the entryPoint is unrecognized the pPrettyName should be housed in symbols to differentiate it from regular contexts (e.g. <entry: main thread>).
Entry points should not be present in any view besides the top-down view (and trace view).
Screenshots
Platform (please complete the following information):
pPrettyName
field. These names are intended for forward-compatibility cases. They are designed to be human-readable but only within the context of an entry point, having them "loose" like this is confusing.To Reproduce Occurs in all multithreaded
meta.db
databases.Expected behavior Entry points should be named with names befitting their context:
entryPoint
field (e.g.(entryPoint: 1)
maps to<program root>
).entryPoint
is unrecognized thepPrettyName
should be housed in symbols to differentiate it from regular contexts (e.g.<entry: main thread>
).Entry points should not be present in any view besides the top-down view (and trace view).
Screenshots
Platform (please complete the following information):
develop
(https://github.com/HPCToolkit/hpcviewer.e4/commit/ade248d3ba182939c472019f81455913b31eb965)