-
Notifications
You must be signed in to change notification settings - Fork 556
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
BBC: Blead breaks SISYPHUS/Math-MPFR-4.21.tar.gz, SISYPHUS/Math-GMPf-0.47.tar.gz, SISYPHUS/Math-Float128-0.15.tar.gz #19550
Comments
@sisyphus FYI |
Bisecting with the following invocation:
... pointed to the following commit:
|
The test is specifically checking the NOK and POK flags on Inf and NaN values, so it is unsurprising that this change results in test failures. Because most of the tests are checking the value of an accumulating count, it also fails on all of those tests after the first true failure: I suspect there are only 2 real test failures here. My guess is that only the tests will need to change, but @sisyphus will need to compare against whatever he's actually trying to achieve here; I'll create an issue for the distro. |
Created sisyphus/math-mpfr#4 |
These tests are checking what the internal flags are after using values in various forms. This is sensitive adjustments to how core handles the flags, which have been intentionally changed in 5.35.10. I'd say that these dists need to have their tests adjusted. The tests are written in a rather strange way, so it's not trivial to propose a fix. I'll let @sisyphus look at fixing them first. |
Sorry I'm a bit late to this - I've been off-grid for a few days. |
The problem is not quite as I initially thought it would be.
On perls prior to the 5.35.10 dev release, this script would output:
But with the release of perl-5.35.10, it outputs:
Do we really want perl's shift() function to alter the flags like that ? However, in my tests I can, of course deal with this bizarre behaviour if I need to. |
I would have assumed not. It's not
This happens in
@nwc10 is this intended behaviour? |
This definitely looks like a bug to me. I believe the code in the SVp_IOK section that disables the SVp_POK flag needs to be copied into the SVp_NOK section. |
Well, it was in 2003 (commit 120fac9) but a lot has changed since then, so I'm not so sure now. But I think that @haarg's suggestion is right. Something like this makes all the modules pass (and the
I'm too tired now to write sensible tests (or a decent commit message), and I'm not sure if there are other places that have been missed. More generally, it might seem inefficient to set the flag bit
|
In the modules of mine that are affected by this, I'll add some code that detects the presence of this behaviour and modifies the failing test script accordingly. AFAICS it's a fairly inconsequential bug. |
Fixed by #19580 (which has not yet been merged at time of writing). |
This (#19580) has now been merged. |
t/NOK_and_POL.t in Math-MPFR-4.21 started to fail with perl 5.35.10:
Math-GMPf-0.47 and Math-Float128-0.15 have a same-named test script, which also started to fail.
The text was updated successfully, but these errors were encountered: