Problem with updating library after removing it from .local/share/fusesoc

[m-kru]

If user removes my_lib directory from .local/share/fusesoc, but leaves entry for that library in fusesoc.conf, and then runs fusesoc update , my_lib is not fetched (is not downloaded to the .local/share/fusesoc). I am not sure if such behavior is desirable, but I personally find it confusing. I think fusesoc update should fetch latest master for all libraries defined in fusesoc.conf file

[olofk]

I think I agree with you. Just trying to think if there are any cases where you don't want to fetch it. Maybe fetch it by default and add a flag that ignores all libraries which aren't already fetched

[m-kru]

In my humble opinion, the behavior of fusesoc library update should be very simple. Namely, update all libraries that are in fusesoc.conf file. By update I mean clone the repository, if it is not in the .local/share/fusesoc directory or pull the latest master if it is already there. If user does not want some library anymore, then delete it from fusesoc.conf (btw. there should probably be some command for this, instead of doing it manually).

[nevercast]

Just encountered this when I depend on https://github.com/davidthings/tinyfpga_bx_usbserial. My project fusesoc.conf is as follows: https://github.com/nevercast/tinysoc/blob/138f1d1def95bff8e5eee929c5e590eb5a3cf4eb/fusesoc.conf

I don't wish to commit fusesoc_libraries/tinyfpga_bx_usbserial to git, but if I remove it then anyone else will face the problem of being unable to build my core.

I expected fusesoc library update to at least reclone it to save me a hot minute, just as described by @m-kru. But I also expected the target that depends on the library to also update it. I can probably live without that last one, but it would be a really nice flow to at least check if libraries are not cloned when a target fails to find a dependency.