> On 7 Feb 2020, at 18:05, Jonathan Gibbons <[email protected]> wrote: > > On 2/7/20 9:22 AM, Pavel Rappo wrote: >> 1. I don't know that code base well, but I'm slowly getting there. The trick >> with this change is to make sure that all those configurations are >> interchangeable (e.g. are the same object). Now, I understand it is*likely* >> to >> be the case. What else could there be? It's just that my eyes started hurting >> when saw all the different ways one could get this object: >> utils.configuration, >> input.utils.configuration, writer.configuration(), through method parameters, >> etc. So I will focus on unraveling that ball of mud. > There is only one configuration per HtmlDoclet instance, and only one > HtmlDoclet > per javadoc invocation. Remember the Highlander movies: "There can only be > one".
Okay, that's what I suspected. I satisfied myself by making sure that the configuration is created only once, that there's no way of copying it, and that nobody seems to clone it. > The configuration is passed around everywhere, but it should always be the > same one. Yeah, the code (not the patch) still leaves me with a feeling of unease. It's as if the objects look around for whatever convenient neighboring objects to ad-hoc provide the required configuration. It should be refactored, though not in this patch. > If you ever see evidence to the contrary, that would be a major, major bug. > > -- Jon Looks good. -Pavel
