It uses the fact that segments are sorted by the linker and '$' as a terminator for the segment name to create unique sorted segment names.
.data$a becomes the start segment.
.data$b is the segment containing the data.
.data$c is the end segment.
The static entries in the start and end segments are zero-sized and thus can be used as start and end pointers.
I know that #2 and the linked pages mentioned issues with name length and padding, but I haven't encountered that in my testing. Maybe they are old details that are no longer relevant?
dtolnay
commented
5 years ago
Works on my machine...
It uses the fact that segments are sorted by the linker and '$' as a terminator for the segment name to create unique sorted segment names.
.data$abecomes the start segment..data$bis the segment containing the data..data$cis the end segment. Thestaticentries in the start and end segments are zero-sized and thus can be used as start and end pointers.I know that #2 and the linked pages mentioned issues with name length and padding, but I haven't encountered that in my testing. Maybe they are old details that are no longer relevant?