Kris Fedorenko
Kris Fedorenko
Sorry, doing a check-in again - looks like this is still an active issue, but it might get addressed in next release (v4.9.1)?
Thank you Ward! Yes I bet there are many things to juggle, thanks for taking a look!
Thanks Dennis! I am not quite clear, what is the expectation for data_read before it goes to nc_free_vlens? The doc just says "pass the pointer back to this function, when...
Oh ok, I see what you mean - there are unused/not-overwritten elements of data_read because of the offsets in both dimensions and only partially writing data to the dimid_y dimension...
Sorry to keep bugging you Dennis, just trying to understand this... I don't think memset is enough, as data_read is an array of nc_vlen_t structs, and memset wouldn't create "legal...
> Memset works because it sets the whole block of memory to zero. Since the nc_vlen_t instances in a vector are contiguous, this ends up filling each nc_vlen_t.len with zeros...
Hi, just wanted to check for any updates! I know this is a very-very edge-case scenario, but we do still see the crash even with the suggested workaround: > More...
Thank you Dennis! It does look potentially related to the mentioned known bug about NC_VLEN and fill values in #2179, although I am not sure what kind of failures were...
It sounds a little bit similar to https://github.com/Unidata/netcdf-c/issues/1985 too...
Hi Dennis! Just wanted to check - is this still an active issue? I know there were a bunch of NC_VLEN fixes in v4.9.0, but I think you mentioned above...