daractive

Yea, that would explain it. Thank you very much.

Thanks for response. Is there any particular reason why it acts like this? I'm mostly interested in knowing why, despite the code looking logical (and the option being there in the first place), it doesn't work. Other write tests seem to handle those data types smoothly.

Basically if I run "test_write_sparse" with all data types supported by version 4 Matlab files (as taken from "Mat_VarWrite4"), only the file for the "double" type seems to work. All other types produce files that Matlab claims to be corrupted and Octave informs that "subscripts must be either integers 1 to (2^63)-1 or logicals". Said error message suggests some error while writing indexes for the array but I can't find connection between data type and said indexes. Attached is a file for a "single"...

Yea, I guess you are right. Either way I think that multiplying by size of data element should solve potential problems, since it would correct potential bugs in padding (since it uses "%8" to determine how much padding bytes are needed) while not affecting success checks (as long as they are just checks against "NULL").

I considered this variable as mostly "funfact" and maybe some debug help, as it affects basically nothing... Except "Mat_VarRead5" where it's actually used for skipping padding in the file which in turn pushed me to ask this question. But AFAIK simply adding multiplications before "fread"s should be enought to actually tweak this one and get rid of issues with skipping padding. Or atleast that's what I deduced from my, still pretty basic, studies of .mat file documentation.

Hello, During combing throught code of this library I noticed something that puzled me. Data reading function (for example "ReadCharData") use variable "bytesread" to store amount of bytes read from a file, to controll padding and the like, but only sometimes it seems to acount for bytes/element, other times just storing raw "fread" return variable (in this case it seems to see no difference between "bytesread" of the same amount of uint8 an uint16 elements). Is this supposed to be like this? Thanks...