Possible memory leak

[cmdcolin]

If you look at a 1000 genome vcf and continuously side scroll, the amount of memory consumed can go up to high levels on the main thread (not web worker). The 1000 genomes vcf's already allocate a lot of memory due to serialization of the genotypes field, but the continuing increase in memory consumption seems odd. It can eventually go down, but the memory remains high and seems like it could be a bug somehow?

It's possible this is not a true memory leak, but i suppose even if not, the amount of memory used for browsing 1000 genomes type data should be brought under control e.g. by avoiding serialization of the genotype field

Example

Memory panel of devtools, showing >2gb consumption, tested in v1.3.0 just for checking if it affects older and newer versions

Allocation sampling panel, showing a large amount of memory dominated by one long shallow rectangle

The code that it points to for that rectangle

I was not sure why that particular bit of code would be "holding" the memory but i'm also not too familiar with memory profiling

[cmdcolin]

share link for the file i'm browsing https://jbrowse.org/code/jb2/v2.4.0/?config=test_data%2Fconfig_demo.json&session=share-j0b0TNXCki&password=LTdhz

[cmdcolin]

might actually be better in newer versions vs the 1.3.0 that i tested

[cmdcolin]

it's really tricky to properly tell if any memory leak exists. wanted to try out https://github.com/plasma-umass/BLeak but couldn't get set up plasma-umass/BLeak#68 (comment)

[cmdcolin]

maybe can close until confirmed