I meant that \documentclass is the first thing in the document _normally_, so reinitializing something that, AFAICT should be initialized and not changed seems to be a no-op. To be clear, I did put \newcount\errorcontextlines before \documentclass{sigplanconf} in my test, and it broke in the same way, with "Ghostscript filter: No bounding box". On Tue, Jun 9, 2015 at 10:17 AM, David Kastrup wrote: > Alex Rozenshteyn writes: > > >> On Tue, Jun 9, 2015 at 10:10 AM, Tassilo Horn wrote: > >> > >>> David Kastrup writes: > >>> > >>> > It might be worth trying to put > >>> > > >>> > \newcount\errorcontextlines > >>> > > >>> > (making the original \errorcontextlines inaccessible) before > >>> > \documentclass in order to see whether this makes a difference. > >>> > >>> No, it doesn't make any difference although I didn't check the preview > >>> output line by line but at least I still get the two errors from above > >>> alternatingly. > > > > Made no difference to me either, though I'd have been rather surprised, > > since \documentclass is the first thing in the document. > > Huh? Both of you realize that I wrote > > put "\newcount\errorcontextlines" *before* \documentclass ? > > Why would it surprise you that this does not make a change "since > \documentclass is the first thing in the document"? \documentclass _is_ > suspected of causing the problem, and so the \newcount\errorcontextlines > would remove the possibility of \documentclass to mess with the setting > of the original \errorcontextlines. For that, it has to come _before_ > \documentclass. > > -- > David Kastrup >