Hi Bart. Thanks for the patches. A few comments On Mon, Apr 03, 2006 at 10:53:28AM -0600, Bart Taylor wrote: > RHEL4-compiler-warnings.patch: > > This is a few quick data type fixes so the source will also compile on > 64 bit RHEL4 without warnings. These changes are on test programs.
The test/ programs do need more attention, it's true. We usually solve them by using the llu() macro, which will cast everything in such a way that both 32 and 64 bit compiles end up clean. > pvfs2-ping.patch: > > Some white space fixes were applied using > pvfs2-mangle-in-place-80col.sh. We've not been too careful with our formatting, have we? However, such whitespace changes make it hard to see the more substantial changes. In the future, separate whitespace changes from all other parts of your patch. > It is also updated to continue after encountering some errors such > as one of the services being unreachable or an unreachable entry > being found in the tab file. This was mainly done by setting flag > if a problem occurs instead of returning -1. Defintely useful. Might not apply it directly, but something like this is a good idea. > pvfs2-genconfig.patch: > > Two options were added and the functionality to run multiple servers > per node. > > The two options --iospecs --metaspecs take arguments of the form > node:port:dir,node:port:dir,?. These are the node name, the port > for the server to run on, and the storage directory for the server > running on that port. This allows any number of IO or meta servers, > each on a unique port and storage location, to run on a node. It > creates the filesystem conf and server conf files appropriately and > moves the log file location entry to the server conf file(s) so that > each service will have its own log. The logfile name always has the > port appended now even in the case with one server or run > interactively. Allowing multiple servers per node is not supported > in the interactive mode. It must be done using the new options. > The ?h option reflects the changes. Sam not that long ago added --ioports and --metaports options so that servers would listen on differetn ports on the same machine and append port number to the storage spaces and log files. Your approach is definitely more flexible, but it seems like we've got an approach that already does what you're trying to do. Are there situations i'm missing where --ioports and --metaports won't do the job? Thanks for sending the patches ==rob -- Rob Latham Mathematics and Computer Science Division A215 0178 EA2D B059 8CDF Argonne National Labs, IL USA B29D F333 664A 4280 315B _______________________________________________ Pvfs2-developers mailing list [email protected] http://www.beowulf-underground.org/mailman/listinfo/pvfs2-developers
