Hi Harald, Pushed as r15-2072. I will wait a few days before backporting but I would be surprised if there are any problems simply because the bug prevented the code patch from doing anything than ICE.
In answer to some of your latest points: > > >> Can we prevent the export of this artificial symbol? > The export of this symbol is needed for interface mapping. Without it, the original bug would reappear. If such len_trim calls are made outside of specification expressions, the symbols are not exported as you verify with either the parse tree dump or direct inspection of the module files from the testcases. > > + char name[2*GFC_MAX_SYMBOL_LEN + 10]; > ^^ this has to be 12 now > > Good spot - thanks for this and all the other help on getting the patch right. Regards Paul
