From: Paul K. <pv...@pv...> - 2013-08-20 12:06:14
|
Christophe Rhodes wrote: > [ I don't have a particularly strong feeling about messages, though I > myself do not plan to act to remove any messages on platforms I don't > use. However... ] Same, but with an emphasis on *remove*. > Gabriel Dos Reis<gd...@in...> writes: >> Remember that despite the failing testsuites, these users are still >> *using* SBCL and they find it useful -- except the banner. If they >> weren't using it, they wouldn't be reporting the annoyance in the >> first place. > > The annoyance, historically, has served two purposes: one, to provide an > unmistakable warning that things might not be as stable as they might be > on this particular platform; two, as a mechanism to force users who are > annoyed by the message to rebuild the system, which provides a minimum > of quality assurance on the platform they are going to be developing > on. While the first purpose is perhaps less appropriate given David's > feeling that there was now nothing fundamentally wrong about the port > (even if the details aren't quite there yet), the second purpose is > still valid: while people aren't reporting (and, ideally, fixing) test > suite problems, a mechanism to encourage users into the thought process > of maintainers is still justifiable. I think it's important to remember the kind of issues that we had when we initially released versions with the CATS message: the initial thread on the win32 port would sometimes return to the C runtime, something that should never happen. I don't remember anyone reporting that they've seen the "CATS ARE NICE." message in a long while, and David's more informed opinion points in the same direction. I also understand that the Windows ports seem particularly weak: their threads are untrustworthy, but there's no way to disable threads either, and the situation is worsened by the fact that few people think of Windows/x86oids as niche platforms. It seems to me the discussion between supporters and opponents of the cats "warning" is running on two parallel tracks: some would like a warning to remain, while others would prefer the "Kitten of Death" message, in all its whimsicality, opacity and compulsoriness to go away. What do you think of: 1. removing the cat messages; 2. replacing the "CATS ARE NICE." lossage with a more informative string; 3. adding a warning in print_banner to the effect that the Windows port may be useful, but is definitely fragile -- one might even add an explicit ask for involvement in development/testing; 4. considering a similar note on other platforms that are the object of less maintenance and testing? In terms of visibility, the only difference between that proposal and the current situation is that the banner can be disabled with --noinform. I'm with Christophe, in that I don't think it's ideal to have a deluge of new users expecting a solid product when we don't have the developers (or even development machines) to respond to bug reports. However, if a project like OpenAxiom ever releases cores with --noinform, I hope it's clear that the onus will be on them to primarily support their end users, even if they hit issues in our code. I would certainly wonder how an inexperienced end-user goes from running Open Axiom to reporting a bug to SBCL, in the absence of the bootup banner (and its associated warning). Paul Khuong |