[darcs-users] Opinion poll for cache handling network options
Christian Kellermann
Christian.Kellermann at nefkom.net
Sun Jan 11 20:46:30 UTC 2009
Dear darcs fans,
I recently wanted to implement a ssh remote command so you can
specify the darcs executable that should get run on the remote end
of a ssh connection. I discovered that none of the usual network
options have any effect. The reason for this is the cache implementation
as it is. It calls copyFileOrUrl with a made up list of DarcsFlags
[] which in my implementation always resulted in the default "darcs"
command.
From my observations I would like to ask for your opinion on how
to proceed here.
1. Flags are ugly, we discussed that on this list already. Jason
has mentioned to me an alternative I would like to investigate in
but I did not bookmark the paper URL and so it is lost for me.
Jason, can you enlighten me on this again? Would investigating a
different mechanism be a worthwile target?
2. Handing through DarcsFlags all over Cache.hs, HashedIO.hs and
friends is awkward. I noticed that some other flags get escaped
with a IORef and are read in via unsafePerformIO. Why that?
3. How should we deal with the current situation? I see the following
bug(s): None of the network options like control master feature are
used with the initial transfer mode. I would expect that since I
can give the option to the commands that invoke transfer-mode. If
this is not considered a bug we need to at least change the misleading
documentation.
Thanks for your insights!
Christian
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 195 bytes
Desc: not available
Url : http://lists.osuosl.org/pipermail/darcs-users/attachments/20090111/f90d4ede/attachment.pgp
More information about the darcs-users
mailing list