On 2 December 2014 at 07:50, Tobias Burnus <bur...@net-b.de> wrote: > Hi Manuel, > > Manuel López-Ibáñez wrote: >> >> This patch actually does not touch linemap but I will appreciate >> Dodji's comments about the approach. The problem is that in case of >> long lines, the column hint of 120 might be too small, thus we do not >> have enough locations within one line to point to a higher column >> (like 132 in the testcase). Giving as column hint the length of the >> line seems the right fix. >> [...] >> My fix here is to create a dummy location for line_len -1 (the last column >> in the line). This forces the next map to start after this last column's >> location. > > > My feeling is that that doesn't work for > -ffree-line-length-none/-ffixed-line-length-none. In that case, > gfc_option.free_line_length == 0 and there is no limit to the number of > characters per line. Thus, I fear that > > b->location > - = linemap_line_start (line_table, current_file->line++, 120); > + = linemap_line_start (line_table, current_file->line++, line_len); > > > might even set the line length to 0 - but I haven't checked. For all other > values, it should work.
My understanding of this piece of code is that Fortran reads the whole line into a buffer, and that line_len is the length of the line just read. Am I wrong? If so, is there any way to compute the length of the current line? A small optimization would be to use all that information about truncation limits to give a smaller column hint when possible, but someone more knowledgeable in Fortran can do that as a follow-up. I think setting it to the actual length line will fix the ICEs and only waste a bit of memory if the lines are quite long. Cheers, Manuel.
