Hi Jerry and Harald, I have on several occasions perused gfc_conv_procedure_call with a view to doing exactly the refactoring that you suggest. However, it has grown like Topsy and the logic has become very difficult to follow. This, of course, makes it ripe for refactoring but has turned it into a substantial task. Thus far I have recoiled with the reaction, "Don't fix it if it (mostly) ain't broke." Something for early 16-branch development?
Regards Paul On Thu, 14 Nov 2024 at 00:18, Jerry D <jvdelis...@gmail.com> wrote: > On 11/13/24 2:26 PM, Harald Anlauf wrote: > > Dear all, > > > > the attached patch is the third part of a series to fix the handling of > > NULL() passed to pointer dummy arguments. This one addresses character > > dummy arguments (scalar, assumed-shape, assumed-rank) for various > > uses in the caller. > > > > The patch is a little larger than I expected, due to corner cases > > (MOLD present or not, assumed-rank or other). If someone finds a > > more clever version, I would be happy to learn about it. > > Especially the treatment of assumed-rank dummy could certainly > > be done differently. > > > > Regtested on x86_64-pc-linux-gnu. OK for mainline? > > > > As this fixes wrong code on the one hand, and is very localized, > > I would like to backport this to 14-branch after some waiting. > > Is this ok? > > > > Thanks, > > Harald > > > > OK for mainline and backport. > > Food for thought at another time: > > gfc_conv_procedure_call contains about 2100 lines of code. One can read > through this fairly directly and the comments are very helpful. It begs > for some refactoring into a set of smaller functions. > > Kind and Type regards, > > Jerry >
