Richard Heck wrote: > On 01/10/2010 10:47 AM, sa...@lyx.org wrote: >> Author: sanda >> Date: Sun Jan 10 16:47:40 2010 >> New Revision: 32942 >> URL: http://www.lyx.org/trac/changeset/32942 >> >> Log: >> This patch from Richard was correct. >> >> > Are you sure? If you don't cancel, then msg is ultimately the return value > from some version of VCS::checkIn(), and all of these seem to return > string() under some circumstances. I can't tell what circumstances, but I'm > not sure we can trust that.
when VCS proceeds we get its result as a log. we parse messages to check what happened (i failed with return codes from svn last time i tried). unless the log is empty we got somewhat transformed log into the nonempty string. so i can imagine there is empty string when eg svn is not found (and reload is not needed either). but right, its very obscure check, and we depend on a very nontransparent machinery, which was meant to be used for statusbar info only. as you say i will add "no message" as a safety measure. my attempt to use setMessage(msg) directly from VCBackend had the problem that another messages like "buffer reloaded" "(vc-check-in)" etc immediately rewrote the status bar. this is not the state in the current trunk because of bug 6417, so for some time window setMessage(msg) is working, but probably get broken again once i fix this bug. pavel
