This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH, ping] Cray Pointers

On Sep 16, 2005, at 7:29 AM, Steve Kargl wrote:

Sigh, gfortran is becoming a cesspool for the bad design decisions
of g77.

The thought has occurred to me more than once. :-(

Alas, it is actually quite difficult to avoid bad design decisions. Users will ask for them, either because the users have codes that depend on them or because the users have different ideas about what good design is (or both). Failing to listen to users is a bad idea, of course. But then, going along with every user request doesn't work either. For a start, there will be directly contradictory user requests.

You will even see users that misunderstand the features their own code is using and thus unintentionally give you incorrect information about what other compilers do. I've seen that one more than once. User says "this works on a Vax". I find old Vax in an obscure corner of the bulding. No it doesn't work. User says, "but my old codes 'just like that' worked". I look at actual old codes. User didn't understand how the features worked and thus incorrectly thought his new stuff was "just like that". Turns out that doing as the user asked would have introduced an incompatibility instead of fixing one. (In particular, user didn't understand the differences in semantics between the standard syntax for PARAMETER and the nonstandard form without the parens. User just thought they were alternate syntax for the same thing; that one is a common misunderstanding, I might add).

Even if you avoid the direct contradictions and misunderstandings, you can get a mess if you go too far down that path. I don't have a magic answer. I don't think there is one. It is a hard problem, which takes considerable thought, along with both the discipline to stick with a development plan, and also an open mind to see when a plan is hurting more than it is helping and needs to be revised. I've seen this kind of quandry on other projects as well.

Dealing with difficult problems like this is no doubt why y'all are paid the big bucks for this work. :-)

Richard Maine                |  Good judgment comes from experience;       |  experience comes from bad judgment.
                            |        -- Mark Twain

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]