From: Thierry L. <thi...@in...> - 2010-01-18 17:31:54
|
Hi, I've just commited the patch we discussed concerning source location information in our branch (including fixes wrt. coding guidelines). Regards, Thierry Lafage. Gabriele Svelto a écrit : > 2009/12/14 Thierry LAFAGE <thi...@in...>: > >> Ok. Thanks. >> I'll try to comply with those quite annoying formatting rules (have to >> better tune my IDE). >> > > Yes, it's the easiest thing to avoid problems, also ensure that the > editor always trims trailing spaces when saving files, there a few > places in GCCs were trailing spaces were left and nobody modified them > for a long time in order not to make merges harder (yeah, it's > completely crazy). > > >> Concerning location information directly from gimple, I hope that this >> information will be correct and accurate in 4.5 since it does not seem to be >> the case at the moment (for this reason I had to pick it into the tree >> representation). If the gimple2cil pass is rewritten to comply with version >> 4.5 then I'll make my code comply to it too. >> > > I have checked in GCC trunk's svn history and apparently > gimple_to_tree() has already been removed so 4.5 won't have the > GIMPLE->TREE->RTL scheme anymore (GIMPLE->TREE->CLI). I was wondering > how they do get proper location information on the other hand because > the TREE from which you extract the source location has been built > from GIMPLE statements. Looking at the old code apparently source > location information was converted from GIMPLE back into the TREE > representation at the bottom of the gimple_to_tree() function in > cfgexpand.c. Here's the relevant snippet: > > /* Set EXPR_LOCATION in all the embedded expressions. */ > loc = gimple_location (stmt); > walk_tree (&t, set_expr_location_r, (void *) &loc, NULL); > > Ciao! > > Gabriele > |