On 11 June 2016 at 19:14, Bart Schaefer wrote: > This is one of the drawbacks to attempting to accurately emulate parsing > rules that are inherently contradictory. It seems that pat="" and then using GLOB_SUBST via $~pat with SH_GLOB temporarily turned off should solve this? But what about (N[1]) etc. modifiers... I'm thinking about a far away possibility that one could fully work with emulate sh, having completions not reporting errors. Best regards, Sebastian Gniazdowski