On Thu, Jun 6, 2019 at 11:43 AM Pavel Sanda <sa...@lyx.org> wrote: > On Wed, Jun 05, 2019 at 07:50:55PM -0600, Joel Kulesza wrote: > > Thanks for putting these together so quickly. Some comments: > > > > 1. p1: The notation seems inconsistent, referring to both citations > and > > references. From what I understand, I would refer strictly to > citations > > (perhaps preceded by bibliography, to help orient a reader). > > I tried to disentangle little bit, see below. > > > 2. p1: I believe that when a citation key cannot be found, the key is > > given. Could this be provided to the user to help him/her better > locate > > the issue? > > Looks little bit difficult within the current machinery (if I understand > it correctly). Cits can be fixed by incremental runs and we see that in > updated result value of scanres of LaTeX::run, i.e. if fixed it does not > contain UNDEF_CIT anymore. > The accompanying structure of errors "terr" does not get reset, so once > you push there some info about missing citation it's not going away. > At the end if you by re-runing fix citation problem but there still remain > some other problem user will wrongly see it within the list of errors... > > I actually do not understand why not reseting "ter" before each run. > Seems reasonable to me, but it's hard to test without having examples > which produced all the trickery in the code. > Otherwise we could just grep in scanLog for the label/citation name > and push it to the error dialog. > > > 3. p2: Perhaps "undefined cross-references"? Same comment about > echoing > > the key that is incorrectly cross-referenced? > > We are getting this error from LaTeX and it says "reference" in both cases > when cit or cross-ref is missing (and maybe in some different cases I do > not > know yet). So I kept it intentionally vague, I changed what we grep for so > it's > hopefully better now. >
I understand. > > > To your concern about being too aggressive: this is potentially an issue. > > However, while I find it a nuisance when I insert a bibliography, haven't > > yet inserted citations, and have to dismiss the errors and "show > anyway," I > > prefer that situation to silent failures that I *don't* expect. The same > > applies here. > > Why I am not particularly confident about the reference part is that some > packages like hyperref might be doing some business with references in > backgrounds (for building ToC or similar) and would flush similar warnings > which are actually harmless in final pdf -- but we will bombard users with > error dialog all the time. > We can try to push to master and wait how much will people scream... > If when this is pushed, I'll try to pull and compile some documents that I suspect will act as good tests. I'm reluctant to now with the patch as it's evolving. > > > Do you have any thoughts on coloring the buttons in the GUI to indicate > > breakage? > > Not enough visibile IMHO, and our current approach seem to be good enough. > Alone, I agree. However, I'd like to see this as a way to help someone track down which citation is a problem as they're working with the document.