Re: RFC: [PATCH hurd 3/6] Add file record locking support: libtrivfs_file_record_lock.patch

[Samuel Thibault]

Svante Signell, le sam. 01 déc. 2018 16:11:32 +0100, a ecrit:
> Found some stuff in IRC logs:

[...]
> 12:13:46< teythoon> gnu_srs: actually, libtrivfs doesn't do the actual node
> handling, it proxies everything to the underlying node

Well, not everything :) notably not the content.

> 12:14:10< teythoon> you should make sure that the locking rpc get's proxied, 
> and
> then libtrivfs is fine i guess

I don't think we want to proxy the record locking, as it's about the
content, which only the translator knows about.  The fact that most
translators open the underlying node with access mode 0 (which entails
the issue you are getting) is a sign of this :)

> 15:43:33< teythoon> gnu_srs: i'm still pretty sure that proxying is the right
> thing to do (if it is for file_chmod, then also for this)

chmod is a different beast, we indeed want the underlying node to
provide the permissions to be used on the translator.

> The updated (non-proxy) implementation now works fine. It does no longer crash
> the null translator :)

Good :)

Samuel