[WIP] Blin 2020.03, round 2

[Altai-man]

• Carp – Fail, Bisected: 7c36816
• Concurrent::Queue – Fail, Bisected: 7ea452c
• Jupyter::Kernel – Fail, Bisected: 40d41a3 (not sure if Blin false positive, but worth checking)
• Lumberjack – Fail, Bisected: 7c36816
• P5caller – Fail, Bisected: 7c36816
• RPi::Device::PiGlow – Fail, Bisected: d8e859d (not sure if false positive, but worth checking)

Ping @vrurg @lizmat

[lizmat]

This fixes P5caller:

-    $down += 3;  # offset heuristic
+    $down += 2;  # offset heuristic

so we're looking at an off-by-one error somewhere...

[jonathanstowe]

The RPi::Device::PiGlow shouldn't even run it's test at all if it's not actually an Raspberry Pi. I'll check when I get indoors.

[jonathanstowe]

With Lumberjack yep I can see how that change might break it. I'll take a look.

[vrurg]

There is something wrong happened with Concurrent::Queue. It tests fine.

[vrurg]

Carp PR: alabamenhu/Carp#3

[vrurg]

Jupyter::Kernel tests ok.

[vrurg]

P5Caller fixed with PR lizmat/P5caller#1

BTW, thanks for the direct links to module sources! It's really convenient!

[vrurg]

@jonathanstowe BTW, look at the submitted PRs for Carp and P5caller. This is the recommended way to maintain the backward compatibility.

[lizmat]

I'm not sure I like the recommendation: we've effectively broken the contract with this change in Backtrace. Changing all of the modules out there (especially the ones we do not know about) is not a solution :-(

[vrurg]

@lizmat Here is a quote from the docs: Its default stringification excludes backtrace frames that are deemed unnecessary or confusing. Yes, it mentions stringification, but since there is no other notion of the default behavior this line can be taken as the desired behavior.

Another confusion comes from t/spec/S32-exceptions/misc.t. At line 460 the test checks for 4 frames, the message talks about 2. Wheres, as a matter of fact, it must be 3 as fail has to be excluded too, but that's off topic.

So, I consider the current behavior of Backtrace correct. Especially considering that previously it was dependent on optimization implementation.

[jonathanstowe]

The RPi::Device::PiGlow shouldn't even run it's test at all if it's not actually an Raspberry Pi. I'll check when I get indoors.

I'm bemused by this failing frankly, the test basically just checks that it compiles as it's impossible to otherwise test without observing the behaviour of the actual lights on the device, so I test by hand before releasing.

[Altai-man]

For the record, blin logs for RPi::Device::PiGlow:

blin@brezeleisen:~$ cat /var/tmp/Blin-volume/Blin/installed/RPi\:\:Device\:\:PiGlow_0.0.3/log-old
The following candidates are already installed: RPi::Device::PiGlow
===> Searching for: RPi::Device::PiGlow
===> Found: RPi::Device::PiGlow:ver<0.0.3>:auth<github:jonathanstowe>:api<1.0> [via Zef::Repository::LocalCache]
===> Testing: RPi::Device::PiGlow:ver<0.0.3>:auth<github:jonathanstowe>:api<1.0>
[RPi::Device::PiGlow] t/001-meta.t .. ok
[RPi::Device::PiGlow] t/010-use.t ... ok
[RPi::Device::PiGlow] All tests successful.
[RPi::Device::PiGlow] Files=2, Tests=2,  2 wallclock secs ( 0.03 usr  0.00 sys +  2.47 cusr  0.22 csys =  2.72 CPU)
[RPi::Device::PiGlow] Result: PASS
===> Testing [OK] for RPi::Device::PiGlow:ver<0.0.3>:auth<github:jonathanstowe>:api<1.0>
blin@brezeleisen:~$ cat /var/tmp/Blin-volume/Blin/installed/RPi\:\:Device\:\:PiGlow_0.0.3/log-new
The following candidates are already installed: RPi::Device::PiGlow
===SORRY!=== Error while compiling /mnt/Blin/installed/RPi::Device::PiGlow_0.0.3/sources/358E929317127004D9B9E2083E9F12DF703DF61E (RPi::Device::PiGlow)
Can never assign default value Str ("/dev/i2c-1") to attribute '$!i2c-bus-device-path', it expects: RPi::Device::SMBus::DevicePath
at /mnt/Blin/installed/RPi::Device::PiGlow_0.0.3/sources/358E929317127004D9B9E2083E9F12DF703DF61E (RPi::Device::PiGlow):187

===> Searching for: RPi::Device::PiGlow
===> Found: RPi::Device::PiGlow:ver<0.0.3>:auth<github:jonathanstowe>:api<1.0> [via Zef::Repository::LocalCache]
===> Testing: RPi::Device::PiGlow:ver<0.0.3>:auth<github:jonathanstowe>:api<1.0>
[RPi::Device::PiGlow] t/001-meta.t .. ok
[RPi::Device::PiGlow] # Failed test 'Can load "RPi::Device::PiGlow" ok'
[RPi::Device::PiGlow] # at t/010-use.t line 7
[RPi::Device::PiGlow] # ===SORRY!=== Error while compiling /mnt/Blin/data/zef-data/store/RPi-Device-PiGlow-0.0.3.tar.gz/RPi-Device-PiGlow-0.0.3/lib/RPi/Device/PiGlow.pm (RPi::Device::PiGlow)
[RPi::Device::PiGlow] # Can never assign default value Str ("/dev/i2c-1") to attribute '$!i2c-bus-device-path', it expects: RPi::Device::SMBus::DevicePath
[RPi::Device::PiGlow] # at /mnt/Blin/data/zef-data/store/RPi-Device-PiGlow-0.0.3.tar.gz/RPi-Device-PiGlow-0.0.3/lib/RPi/Device/PiGlow.pm (RPi::Device::PiGlow):187
[RPi::Device::PiGlow] # 
[RPi::Device::PiGlow] # You failed 1 test of 1
[RPi::Device::PiGlow] t/010-use.t ... 
[RPi::Device::PiGlow] Dubious, test returned 1 (wstat 256, 0x100)
[RPi::Device::PiGlow] Failed 1/1 subtests 
[RPi::Device::PiGlow] Test Summary Report
[RPi::Device::PiGlow] -------------------
[RPi::Device::PiGlow] t/010-use.t (Wstat: 256 Tests: 1 Failed: 1)
[RPi::Device::PiGlow]   Failed test:  1
[RPi::Device::PiGlow]   Non-zero exit status: 1
[RPi::Device::PiGlow] Files=2, Tests=2,  3 wallclock secs ( 0.05 usr  0.00 sys +  5.29 cusr  0.30 csys =  5.64 CPU)
[RPi::Device::PiGlow] Result: FAIL
===> Testing [FAIL]: RPi::Device::PiGlow:ver<0.0.3>:auth<github:jonathanstowe>:api<1.0>
[RPi::Device::PiGlow] Failed to get passing tests, but continuing with --force-test
===> Installing: RPi::Device::PiGlow:ver<0.0.3>:auth<github:jonathanstowe>:api<1.0>
===> Install [FAIL] for RPi::Device::PiGlow:ver<0.0.3>:auth<github:jonathanstowe>:api<1.0>: ===SORRY!=== Error while compiling /mnt/Blin/installed/RPi::Device::PiGlow_0.0.3/sources/358E929317127004D9B9E2083E9F12DF703DF61E (RPi::Device::PiGlow)
Can never assign default value Str ("/dev/i2c-1") to attribute '$!i2c-bus-device-path', it expects: RPi::Device::SMBus::DevicePath
at /mnt/Blin/installed/RPi::Device::PiGlow_0.0.3/sources/358E929317127004D9B9E2083E9F12DF703DF61E (RPi::Device::PiGlow):187

blin@brezeleisen:~$ 

[Altai-man]

@lizmat I think the rule of the thumb is: 1)If that's a fix of buggy behavior others has relied upon, better to fix that, even with consequences; 2)If it is not a technical bug fix, but a change for a change, do smarter things to save modules from anything.
@vrurg is it 1 or 2?

[vrurg]

@Altai-man It's 1 since it passes the existing spectests. Especially considering that Backtrace is now deterministic and the test itself doesn't depend on its location in the file anymore.

[jonathanstowe]

[RPi::Device::PiGlow] # Can never assign default value Str ("/dev/i2c-1") to attribute '$!i2c-bus-device-path', it expects: RPi::Device::SMBus::DevicePath
[RPi::Device::PiGlow] # at /mnt/Blin/data/zef-data/store/RPi-Device-PiGlow-0.0.3.tar.gz/RPi-Device-PiGlow-0.0.3/lib/RPi/Device/PiGlow.pm (RPi::Device::PiGlow):187
[RPi::Device::PiGlow] #

Ah yes, it's a subset that checks that the device exists, which it won't on anything but an actualy RPi, I'd just ignore it for the time being while I make some adjustments so the default is done a bit more sanely.

[jonathanstowe]

Fixed the Lumberjack with the jonathanstowe/Lumberjack@0173711 will release later.

[jonathanstowe]

Fixed the Lumberjack with the jonathanstowe/Lumberjack@0173711 will release later.

Made release of v0.0.10 and tested on both old and new rakudo.

[Altai-man]

Marking the Backtrace thing as a serious change to note in changelog.
Thanks to everyone! Another round next weekend.