profiling.sampling heatmap reads non-regular and oversized files

[tonghuaroot]

Bug description

The heatmap exporter in profiling.sampling reads source files named in a profile unconditionally via Path(filename).read_text(). It does not check whether the path is a regular file or whether the file size is reasonable.

When replaying a binary profile, a filename pointing to a device file (e.g. /dev/zero) causes unbounded memory growth. An oversized file has the same effect.

The fix adds an is_file() guard and a 10 MB size cap before reading, consistent with the binary reader hardening in gh-152718.

CPython versions tested on

main

Operating systems tested on

macOS

Linked PRs

• gh-153838: Skip non-regular source files in the heatmap exporter #153839
• [3.15] gh-153838: Skip non-regular source files in the heatmap exporter (GH-153839) #158853

[picnixz]

I do not think it is correct. Files may be very huge for very long apps. It is different from the binary file read protection

[tonghuaroot]

Good point. Dropped the size cap, kept the is_file() guard only.