On Wed, Sep 07, 2016 at 05:33:08PM +0800, Fam Zheng wrote: > On Wed, 09/07 10:23, Alex Bennée wrote: > > Last time I needed to do error reporting I was told the error_setg > > method was the correct way to do it and the report/exit case made sense > > only in the top level. Before we add even more error reporting > > primitives can we update HACKING (or possibly add somethings to docs/) > > that describes when and where the various error report mechanisms should > > be used? > > FWIW, the comment of error_setg clarifies the recommended usage now: > > ... > > * Please don't error_setg(&error_fatal, ...), use error_report() and > > * exit(), because that's more obvious. > > * Likewise, don't error_setg(&error_abort, ...), use assert(). > > */ > > #define error_setg(errp, fmt, ...) \
I can update both HACKING and above comment to mention about the new interface in v5 (if we will have a v5) by appending a new patch for the docs. Otherwise, I can post the patch seperately when needed. Thanks, -- peterx
