On Thu, Jun 23, 2011 at 1:17 PM, Julian Foad <julian.f...@wandisco.com> wrote: > Greg Stein wrote: >> Thanks for taking the time to do thus. Only recently did I realize that I >> named the function incorrectly when I came up with it. In PoCore, I called >> it pc_error_trace(), for the same reasons you suggested here. > > OK, I'll do it. > > Does anyone want me to leave "svn_error_return" as an alias for the time > being, to be removed before 1.7.0, to avoid gratuitous failures with > developers' local mods and when applying patches over the next few days? > I'll happily do so if anyone says "yes" (it's very little extra work), > but if no-one says "yes" then I won't.
Don't bother leaving the alias. Folks will have to fix it sooner or later, might as well have them do it sooner. >> > There are 695 uses of svn_error_return. >> > >> > 681 of them are "return svn_error_return(...);". >> > >> > The other 14 are "some_err = svn_error_return(...);". >> > >> > So if we're going to change the vast majority of these in any way, what >> > do folks feel about replacing them with >> > >> > SVN_ERR_RETURN(...) >> >> No. >> >> SVN_ERR() is conditional. Code typically occurs afterwards, and is executed >> in the non-exceptional case. You're talking about creating a normally-used >> macro that will hide this control flow change. > > OK, good to hear your input. Same. I knew there was a reason we didn't use this form when we were designing this macro, but I just couldn't put my finger on it. -Hyrum
