On Sun, Mar 23, 2025 at 02:44:29PM -0500, Robert Dubner wrote: > I am enclosing a patch to be applied on top of yours. (Your patch got us > down to zero errors in the "Coughlan" tests, 2 UAT errors, and 4 errors in > the NIST tests. Well done!) > > This one passes all of my tests, in both ASCII and EBCDIC forms. It also > passes "make check-cobol". That's on my x_86_64-linux machine. > > (That's the good news. The bad news is that this is exposing gaps in > coverage of our test suites. There is stuff that the misnamed numstr2i > routine used to do that isn't being done, but no errors are flagged in any > test.) > > Given that this version passes everything that our regression tests cover, > is it time to accumulate all this work into a single patch and have that > committed? > > Perhaps I should create that patch, seeing as how at this moment only I > can do all of my known tests.
Note, at least on my side, I haven't yet groked the formatting rules in the COBOL FE, so feel free to reindent/reformat it in the style you're used to. > diff --git a/gcc/cobol/genapi.cc b/gcc/cobol/genapi.cc > index ca86ffa2fc74..f3cab0a4ad1e 100644 > --- a/gcc/cobol/genapi.cc > +++ b/gcc/cobol/genapi.cc > @@ -4897,8 +4897,7 @@ parser_display_internal(tree file_descriptor, > } > else > { > - p += 1; > - int exp = atoi(p); > + int exp = atoi(p+1); > if( exp >= 6 || exp <= -5 ) > { > // We are going to stick with the E notation, so ach has our > result Oops, yes, without it all the e -> E and exponent rewriting is wrong. I wrote that part without actually testing it on anything. Jakub