Can I delete the protobuf jar that comes with recent Matlab installations?

The protobuf jars included with Matlab installations starting with 2020a (and perhaps 2019b) have conflicting classes and effectively make it so that Matlab is incapable of parsing protobuf messages. I've submitted bug reports and posted a question about how to work around this issue. One possible work around is to delete the protobuf jar that is included with the installation or otherwise adjust the static Java path, which will allow my custom protobuf jar, added to the dynamic Java path, to take precedence. I've resisted that because among other things I don't want to break anything going on under the hood. My question here is: what effect will that have on internal operations? What internal functionality is that jar controlling and how might it change if I use my own jar? Thanks!

Answer 1

[Muskan]

Hi Matt,

As per my understanding, I would suggest against deleting the protobuf jar that comes with MATLAB installations. The jar files contain the protobuf library, which allows MATLAB to read and write protobuf data.

You can experience problems using MATLAB features that rely on protobuf if you delete protobuf jar. Moreover, deleting system files without understanding their function may result in system instability.

Cheers,

Muskan

[Matt]

Hi Muskan,

The protobuf jars actually prevent MATLAB from reading and writing protobuf data; that is the entire problem I'm having. I submitted a bug report a while back with the details but I can't recall them at the moment, other than that there are conflicting versions with the classes within the included protobuf jar. If I recall, the guava version inside the included jar is very old (released sometime around 2010) and is incompatible with other classes within that jar, so if you have written protobuf parsing code for your own messages that relies on guava, it errors out. Since we need guava to parse our messages, we can't use versions of Matlab that have an included protobuf jar and are stuck on 2019a. I was hoping the issue had been addressed with newer versions of Matlab, but as of 2022a it was still an issue.

Thanks,

Matt