Skip Menu |
Report information
Id: 131735
Status: resolved
Priority: 0/
Queue: perl5

Owner: jkeenan <jkeenan [at] cpan.org>
Requestors: carlos [at] carlosguevara.com
Cc:
AdminCc:

Operating System: netbsd
PatchStatus: (no value)
Severity: low
Type: unknown
Perl Version: (no value)
Fixed In: 5.27.3



Date: Tue, 11 Jul 2017 01:30:36 -0500
From: Carlos Guevara <carlos [...] carlosguevara.com>
Subject: NetBSD 7.1 Smoke Test Failures
To: perlbug [...] perl.org
Download (untitled) / with headers
text/plain 426b
Attached file netbsd.failures.txt contains the results of the following on NetBSD 7.1:
sh ./Configure -des -Dusedevel -Duseithreads -Duse64bitint
make test_prep
cd t; ./perl harness -v ../cpan/Sys-Syslog/t/syslog.t ../lib/locale.t 2>&1 | tee ../netbsd.failures.txt; cd -

Attached file perl-Ilib-V.txt contains the results of './perl -Ilib -V' in the same directory.

Please let me know if I can be of any further assistance.

Download netbsd.failures.txt
text/plain 58.7k

Message body is not shown because sender requested not to inline it.

Download perl-Ilib-V.txt
text/plain 2.5k

Message body is not shown because sender requested not to inline it.

RT-Send-CC: perl5-porters [...] perl.org
Download (untitled) / with headers
text/plain 1.1k
On Tue, 11 Jul 2017 06:31:33 GMT, carlos@carlosguevara.com wrote: Show quoted text
> Attached file netbsd.failures.txt contains the results of the following on > NetBSD 7.1: > sh ./Configure -des -Dusedevel -Duseithreads -Duse64bitint > make test_prep > cd t; ./perl harness -v ../cpan/Sys-Syslog/t/syslog.t ../lib/locale.t 2>&1 > | tee ../netbsd.failures.txt; cd - > > Attached file perl-Ilib-V.txt contains the results of './perl -Ilib -V' in > the same directory. > > Please let me know if I can be of any further assistance.
Thanks for preparing that report. In gratitude, we ask you to prepare more reports! The failures in ../cpan/Sys-Syslog/t/syslog.t are not consistent from one smoke run to the next. However, they are like those being newly reported for DragonflyBSD as well. Let's set those aside for now. The output from lib/locale.t suggests how to proceed. In the same directory, could you run the following command and include the output as an attachment in responding to this message? ##### cd t; PERL_DEBUG_FULL_TEST=2 ./perl harness -v ../lib/locale.t 2>&1 | gzip -c > ../lib-locale-t-netbsd.txt.gz; cd - ##### Thank you very much. -- James E Keenan (jkeenan@cpan.org)
Date: Thu, 13 Jul 2017 00:54:56 -0500
From: Carlos Guevara <carlos [...] carlosguevara.com>
Subject: Re: [perl #131735] NetBSD 7.1 Smoke Test Failures
To: perlbug-followup [...] perl.org
Download (untitled) / with headers
text/plain 1.3k
No problem.

Report attached.


On Wed, Jul 12, 2017 at 7:59 AM, James E Keenan via RT <perlbug-followup@perl.org> wrote:
Show quoted text
On Tue, 11 Jul 2017 06:31:33 GMT, carlos@carlosguevara.com wrote:
> Attached file netbsd.failures.txt contains the results of the following on
> NetBSD 7.1:
> sh ./Configure -des -Dusedevel -Duseithreads -Duse64bitint
> make test_prep
> cd t; ./perl harness -v ../cpan/Sys-Syslog/t/syslog.t ../lib/locale.t 2>&1
> | tee ../netbsd.failures.txt; cd -
>
> Attached file perl-Ilib-V.txt contains the results of './perl -Ilib -V' in
> the same directory.
>
> Please let me know if I can be of any further assistance.

Thanks for preparing that report.  In gratitude, we ask you to prepare more reports!

The failures in ../cpan/Sys-Syslog/t/syslog.t are not consistent from one smoke run to the next.  However, they are like those being newly reported for DragonflyBSD as well.  Let's set those aside for now.

The output from lib/locale.t suggests how to proceed.  In the same directory, could you run the following command and include the output as an attachment in responding to this message?

#####
cd t; PERL_DEBUG_FULL_TEST=2 ./perl harness -v ../lib/locale.t 2>&1 | gzip -c > ../lib-locale-t-netbsd.txt.gz; cd -
#####

Thank you very much.

--
James E Keenan (jkeenan@cpan.org)

Download lib-locale-t-netbsd.txt.gz
application/x-gzip 41k

Message body not shown because it is not plain text.

RT-Send-CC: perl5-porters [...] perl.org
Download (untitled) / with headers
text/plain 470b
On Thu, 13 Jul 2017 13:27:35 GMT, carlos@carlosguevara.com wrote: Show quoted text
> No problem. > > Report attached. >
Thanks. Last year Karl Williamson used output like this to diagnose and correct failures in lib/locale.t on FreeBSD. I'm hoping he'll be able to do likewise on NetBSD. If you can continue to submit smoke test reports for Perl 5 blead on NetBSD 7 -- say, weekly or semi-weekly -- that would be helpful. Thank you very much. -- James E Keenan (jkeenan@cpan.org)
Date: Thu, 13 Jul 2017 12:59:21 -0600
From: Karl Williamson <public [...] khwilliamson.com>
To: Carlos Guevara <carlos [...] carlosguevara.com>, perlbug-followup [...] perl.org
Subject: Re: [perl #131735] NetBSD 7.1 Smoke Test Failures
Download (untitled) / with headers
text/plain 5.7k
On 07/12/2017 11:54 PM, Carlos Guevara wrote: Show quoted text
> No problem. > > Report attached. > > > On Wed, Jul 12, 2017 at 7:59 AM, James E Keenan via RT > <perlbug-followup@perl.org <mailto:perlbug-followup@perl.org>> wrote: > > On Tue, 11 Jul 2017 06:31:33 GMT, carlos@carlosguevara.com > <mailto:carlos@carlosguevara.com> wrote:
> > Attached file netbsd.failures.txt contains the results of the
> following on
> > NetBSD 7.1: > > sh ./Configure -des -Dusedevel -Duseithreads -Duse64bitint > > make test_prep > > cd t; ./perl harness -v ../cpan/Sys-Syslog/t/syslog.t
> ../lib/locale.t 2>&1
> > | tee ../netbsd.failures.txt; cd - > > > > Attached file perl-Ilib-V.txt contains the results of './perl
> -Ilib -V' in
> > the same directory. > > > > Please let me know if I can be of any further assistance.
> > Thanks for preparing that report. In gratitude, we ask you to > prepare more reports! > > The failures in ../cpan/Sys-Syslog/t/syslog.t are not consistent > from one smoke run to the next. However, they are like those being > newly reported for DragonflyBSD as well. Let's set those aside for now. > > The output from lib/locale.t suggests how to proceed. In the same > directory, could you run the following command and include the > output as an attachment in responding to this message? > > ##### > cd t; PERL_DEBUG_FULL_TEST=2 ./perl harness -v ../lib/locale.t 2>&1 > | gzip -c > ../lib-locale-t-netbsd.txt.gz; cd - > ##### > > Thank you very much. > > -- > James E Keenan (jkeenan@cpan.org <mailto:jkeenan@cpan.org>) > >
This output indicates problems with the NetBSD locale definitions, rather than Perl. It would be good if someone, not me, would report them to the vendor. Other platforms also have various defective locales, which we can tell lib/locale.t about so that it doesn't fail. I have seen almost all these failures on other platforms, but some are new to me. lib/locale.t is written to look for anomalies in the locale definitions. It does not catch everything, but quite a few. I basically looked through the POSIX standard and wrote tests for just about everything I could think of and that I could figure out how to test, that it specifies. For example all lowercase characters should also be alphabetic, and none should be digits nor punctuation. You can see the failures by grepping for 'failed' in the output. They follow, grouped by locale, with my comments after # failed 431 (Verify that isn't both [:graph:] and [:space:]) with locale 'cs_CZ.ISO8859-2' (for A0) # failed 431 (Verify that isn't both [:graph:] and [:space:]) with locale 'hr_HR.ISO8859-2' (for A0) # failed 431 (Verify that isn't both [:graph:] and [:space:]) with locale 'hu_HU.ISO8859-2' (for A0) # failed 431 (Verify that isn't both [:graph:] and [:space:]) with locale 'lt_LT.ISO8859-4' (for A0) # failed 431 (Verify that isn't both [:graph:] and [:space:]) with locale 'pl_PL.ISO8859-2' (for A0) # failed 431 (Verify that isn't both [:graph:] and [:space:]) with locale 'ro_RO.ISO8859-2' (for A0) # failed 431 (Verify that isn't both [:graph:] and [:space:]) with locale 'sk_SK.ISO8859-2' (for A0) # failed 431 (Verify that isn't both [:graph:] and [:space:]) with locale 'sl_SI.ISO8859-2' (for A0) # failed 431 (Verify that isn't both [:graph:] and [:space:]) with locale 'sr_Latn_ME.ISO8859-2' (for A0) # failed 431 (Verify that isn't both [:graph:] and [:space:]) with locale 'sr_Latn_RS.ISO8859-2' (for A0) # failed 431 (Verify that isn't both [:graph:] and [:space:]) with locale 'sr_ME.ISO8859-2' (for A0) # failed 431 (Verify that isn't both [:graph:] and [:space:]) with locale 'sr_RS.ISO8859-2' (for A0) 0xA0 is the No-break space in ISO 8859-2. Posix stipulates that something can't both be a space character and a graphic one. (Instead spaces are printable) This is a defective locale definition. AIX has this all over the place, so we suppress the output of these tests there. # failed 398 (Verify that /[[:upper:]]/ matches all alpha X for which uc(X) == X and lc(X) != X) with locale 'lt_LT.ISO8859-4' (for BD) # failed 399 (Verify that /[[:lower:]]/i matches all alpha X for which uc(X) == X and lc(X) != X) with locale 'lt_LT.ISO8859-4' (for BD) # failed 465 (Verify case insensitive matching works) with locale 'lt_LT.ISO8859-4' (for BD BF) BD and BF are upper and lowercase "Ŋ" (Eng) characters in 8859-4. The locale definition is not correctly saying that these are a case pair. We have seen this before, I believe on Darwin. # failed 407 (Verify that [:lower:] is a subset of [:alpha:]) with locale 'lt_LT.ISO8859-13' (for DF) # failed 410 (Verify that /[[:lower:]]/i is a subset of [:alpha:]) with locale 'lt_LT.ISO8859-13' (for DF) DF is the German "ß" alphabetic letter. This is a defect in the locale definition, as it definitely is both :lower: and :alpha:. # failed 426 (Verify that the space character is in [:print:]) with locale 'kk_KZ.PT154' (for 20) This is the one I've never seen before. I was unfamiliar with this locale, which is for Kazakh. Perl works only on locales that are essentially supersets of ASCII. So for example, it expects that a SPACE is 0x20 (or 0x40 on EBCDIC platforms). Imagine how bad the results would be if a regular expression pattern looking for a SPACE, can't ever find one because the locale defines it differently, and instead finds non-spaces. I don't think there is any way to test that the code point at that position in the locale is SPACE, but I can test that it is a member of the :space: POSIX class (and that test passed), and that is also a member of :print: I looked the locale up online, and 0x20 it does appear to be a superset of ASCII, with SPACE at 0x20. So this is a defective locale definition.
Date: Fri, 14 Jul 2017 11:19:54 -0600
From: Karl Williamson <public [...] khwilliamson.com>
To: Carlos Guevara <carlos [...] carlosguevara.com>, perlbug-followup [...] perl.org
Subject: Re: [perl #131735] NetBSD 7.1 Smoke Test Failures
Download (untitled) / with headers
text/plain 2.1k
On 07/13/2017 12:59 PM, Karl Williamson wrote: Show quoted text
> On 07/12/2017 11:54 PM, Carlos Guevara wrote:
>> No problem. >> >> Report attached. >> >> >> On Wed, Jul 12, 2017 at 7:59 AM, James E Keenan via RT >> <perlbug-followup@perl.org <mailto:perlbug-followup@perl.org>> wrote: >> >> On Tue, 11 Jul 2017 06:31:33 GMT, carlos@carlosguevara.com >> <mailto:carlos@carlosguevara.com> wrote:
>> > Attached file netbsd.failures.txt contains the results of the
>> following on
>> > NetBSD 7.1: >> > sh ./Configure -des -Dusedevel -Duseithreads -Duse64bitint >> > make test_prep >> > cd t; ./perl harness -v ../cpan/Sys-Syslog/t/syslog.t
>> ../lib/locale.t 2>&1
>> > | tee ../netbsd.failures.txt; cd - >> > >> > Attached file perl-Ilib-V.txt contains the results of './perl
>> -Ilib -V' in
>> > the same directory. >> > >> > Please let me know if I can be of any further assistance.
>> >> Thanks for preparing that report. In gratitude, we ask you to >> prepare more reports! >> >> The failures in ../cpan/Sys-Syslog/t/syslog.t are not consistent >> from one smoke run to the next. However, they are like those being >> newly reported for DragonflyBSD as well. Let's set those aside >> for now. >> >> The output from lib/locale.t suggests how to proceed. In the same >> directory, could you run the following command and include the >> output as an attachment in responding to this message? >> >> ##### >> cd t; PERL_DEBUG_FULL_TEST=2 ./perl harness -v ../lib/locale.t 2>&1 >> | gzip -c > ../lib-locale-t-netbsd.txt.gz; cd - >> ##### >> >> Thank you very much. >> >> -- >> James E Keenan (jkeenan@cpan.org <mailto:jkeenan@cpan.org>) >> >>
> > This output indicates problems with the NetBSD locale definitions, > rather than Perl. It would be good if someone, not me, would report > them to the vendor. >
But there are indications in the output you gave us that there are bugs in the test script. Please run the attached script instead. You can put it in the /lib directory, and run the command the Jim gave you, using whatever name you chose instead of locale.t
Download locale.t
text/plain 95.4k

Message body is not shown because sender requested not to inline it.

Date: Sun, 16 Jul 2017 11:58:13 -0500
From: Carlos Guevara <carlos [...] carlosguevara.com>
CC: perlbug-followup [...] perl.org
Subject: Re: [perl #131735] NetBSD 7.1 Smoke Test Failures
To: Karl Williamson <public [...] khwilliamson.com>
Download (untitled) / with headers
text/plain 2.3k
No problem.

Reports attached.


On Fri, Jul 14, 2017 at 12:19 PM, Karl Williamson <public@khwilliamson.com> wrote:
Show quoted text
On 07/13/2017 12:59 PM, Karl Williamson wrote:
On 07/12/2017 11:54 PM, Carlos Guevara wrote:
No problem.

Report attached.


On Wed, Jul 12, 2017 at 7:59 AM, James E Keenan via RT <perlbug-followup@perl.org <mailto:perlbug-followup@perl.org>> wrote:

    On Tue, 11 Jul 2017 06:31:33 GMT, carlos@carlosguevara.com
    <mailto:carlos@carlosguevara.com> wrote:
     > Attached file netbsd.failures.txt contains the results of the
    following on
     > NetBSD 7.1:
     > sh ./Configure -des -Dusedevel -Duseithreads -Duse64bitint
     > make test_prep
     > cd t; ./perl harness -v ../cpan/Sys-Syslog/t/syslog.t
    ../lib/locale.t 2>&1
     > | tee ../netbsd.failures.txt; cd -
     >
     > Attached file perl-Ilib-V.txt contains the results of './perl
    -Ilib -V' in
     > the same directory.
     >
     > Please let me know if I can be of any further assistance.

    Thanks for preparing that report.  In gratitude, we ask you to
    prepare more reports!

    The failures in ../cpan/Sys-Syslog/t/syslog.t are not consistent
    from one smoke run to the next.  However, they are like those being
    newly reported for DragonflyBSD as well.  Let's set those aside for now.

    The output from lib/locale.t suggests how to proceed.  In the same
    directory, could you run the following command and include the
    output as an attachment in responding to this message?

    #####
    cd t; PERL_DEBUG_FULL_TEST=2 ./perl harness -v ../lib/locale.t 2>&1
    | gzip -c > ../lib-locale-t-netbsd.txt.gz; cd -
    #####

    Thank you very much.

    --
    James E Keenan (jkeenan@cpan.org <mailto:jkeenan@cpan.org>)



This output indicates problems with the NetBSD locale definitions, rather than Perl.  It would be good if someone, not me, would report them to the vendor.


But there are indications in the output you gave us that there are bugs in the test script.  Please run the attached script instead.  You can put it in the /lib directory, and run the command the Jim gave you, using whatever name you chose instead of locale.t

Download perl-Ilib-V.txt
text/plain 2.5k

Message body is not shown because sender requested not to inline it.

Message body not shown because it is not plain text.

Subject: Re: [perl #131735] NetBSD 7.1 Smoke Test Failures
To: Carlos Guevara <carlos [...] carlosguevara.com>
From: Karl Williamson <public [...] khwilliamson.com>
Date: Mon, 17 Jul 2017 14:43:56 -0600
CC: perlbug-followup [...] perl.org
Download (untitled) / with headers
text/plain 5.8k
On 07/16/2017 10:58 AM, Carlos Guevara wrote: Show quoted text
> No problem. > > Reports attached. >
Thanks. That showed me that the previous version of lib/locale.t had a bug and was not correctly reporting some failures. And the updated version of that script, which I sent you to try, actually fixed it. The new script is now in blead, commit 6b88b15be98ae595ce9dc9045df48ed656d073be. Fixing that also correctly showed why netbsd is failing this test. 8% of the locales on your system, all of them being the ISO 8859-2 type, say that a No-break space (NBSP) character is graphic. A graphic character is defined as something that would use ink when being printed, but a NBSP is just empty space with the special property that the words on either side of it shouldn't be split across lines when being output to, say, a display. Because many systems have similar defective locales, an error isn't actually raised unless a threshold of failing tests is reached. This is to prevent us from having to deal with failing smokes that aren't really Perl bugs. But at some fairly small threshold of failure, it could be a Perl bug that we should look at. What I decided rather arbitrarily, is if 95% or more of the locales pass an individual test in this script, then the failure is more likely to be in the failing locales than it is to be in Perl, and so the test passes. But in your system, only 92% of all the locales pass this one test. It turns out that it is a bug in your locales, but the script did its job and notified us of a potential problem. On other systems, I've ended up adjusting the threshold. On AIX machines, almost all the locales have the same defect as your 8859-2 ones do. The threshold there has been jacked up to if only 0.1% pass, the script doesn't fail. (Note that your 8859-1 locales don't have the same problem, and yet they have the NBSP character same as the 8859-2 ones. I wonder if the definition of 8859-2 changed at some point, and these locale definitions didn't get fully updated. Or maybe they all come from the same original definition, and the person who wrote 8859-2 differed from the one who wrote 8859-1, and wasn't as clued in, or they just were having a bad day, or ...) Some of your other locales also are being found to have defects by this test script, but in rates well below 5% of all the locales, so this is the only failure that gets raised. (The other failures are shown if you run the script by hand, but they are marked in such a way as to prevent them from being shown as failures by the overall process.) I'll think about the appropriate way to change the test to not cause this script to be marked as failing. But these are real defects that I believe should be reported to the OS vendor. Show quoted text
> > On Fri, Jul 14, 2017 at 12:19 PM, Karl Williamson > <public@khwilliamson.com <mailto:public@khwilliamson.com>> wrote: > > On 07/13/2017 12:59 PM, Karl Williamson wrote: > > On 07/12/2017 11:54 PM, Carlos Guevara wrote: > > No problem. > > Report attached. > > > On Wed, Jul 12, 2017 at 7:59 AM, James E Keenan via RT > <perlbug-followup@perl.org > <mailto:perlbug-followup@perl.org> > <mailto:perlbug-followup@perl.org > <mailto:perlbug-followup@perl.org>>> wrote: > > On Tue, 11 Jul 2017 06:31:33 GMT, > carlos@carlosguevara.com <mailto:carlos@carlosguevara.com> > <mailto:carlos@carlosguevara.com > <mailto:carlos@carlosguevara.com>> wrote:
> > Attached file netbsd.failures.txt contains the
> results of the > following on
> > NetBSD 7.1: > > sh ./Configure -des -Dusedevel -Duseithreads
> -Duse64bitint
> > make test_prep > > cd t; ./perl harness -v ../cpan/Sys-Syslog/t/syslog.t
> ../lib/locale.t 2>&1
> > | tee ../netbsd.failures.txt; cd - > > > > Attached file perl-Ilib-V.txt contains the results
> of './perl > -Ilib -V' in
> > the same directory. > > > > Please let me know if I can be of any further
> assistance. > > Thanks for preparing that report. In gratitude, we ask > you to > prepare more reports! > > The failures in ../cpan/Sys-Syslog/t/syslog.t are not > consistent > from one smoke run to the next. However, they are like > those being > newly reported for DragonflyBSD as well. Let's set > those aside for now. > > The output from lib/locale.t suggests how to proceed. > In the same > directory, could you run the following command and > include the > output as an attachment in responding to this message? > > ##### > cd t; PERL_DEBUG_FULL_TEST=2 ./perl harness -v > ../lib/locale.t 2>&1 > | gzip -c > ../lib-locale-t-netbsd.txt.gz; cd - > ##### > > Thank you very much. > > -- > James E Keenan (jkeenan@cpan.org > <mailto:jkeenan@cpan.org> <mailto:jkeenan@cpan.org > <mailto:jkeenan@cpan.org>>) > > > > This output indicates problems with the NetBSD locale > definitions, rather than Perl. It would be good if someone, not > me, would report them to the vendor. > > > But there are indications in the output you gave us that there are > bugs in the test script. Please run the attached script instead. > You can put it in the /lib directory, and run the command the Jim > gave you, using whatever name you chose instead of locale.t > >
Subject: Re: [perl #131735] NetBSD 7.1 Smoke Test Failures
To: Carlos Guevara <carlos [...] carlosguevara.com>
From: Karl Williamson <public [...] khwilliamson.com>
Date: Mon, 17 Jul 2017 15:40:24 -0600
CC: perlbug-followup [...] perl.org
Download (untitled) / with headers
text/plain 668b
On 07/17/2017 02:43 PM, Karl Williamson wrote: Show quoted text
> Fixing that also correctly showed why netbsd is failing this test. 8% > of the locales on your system, all of them being the ISO 8859-2 type, > say that a No-break space (NBSP) character is graphic. A graphic > character is defined as something that would use ink when being printed, > but a NBSP is just empty space with the special property that the words > on either side of it shouldn't be split across lines when being output > to, say, a display.
After a couple more minutes of thought (less than I thought would be necessary), I think I have a solution. Please run the attached script, like before.
Download locale.t
text/plain 95.3k

Message body is not shown because sender requested not to inline it.

RT-Send-CC: perl5-porters [...] perl.org
Download (untitled) / with headers
text/plain 3.5k
On Mon, 17 Jul 2017 20:46:07 GMT, public@khwilliamson.com wrote: Show quoted text
> On 07/16/2017 10:58 AM, Carlos Guevara wrote:
> > No problem. > > > > Reports attached. > >
> > Thanks. > > That showed me that the previous version of lib/locale.t had a bug and > was not correctly reporting some failures. And the updated version of > that script, which I sent you to try, actually fixed it. The new script > is now in blead, commit 6b88b15be98ae595ce9dc9045df48ed656d073be. > > Fixing that also correctly showed why netbsd is failing this test. 8% > of the locales on your system, all of them being the ISO 8859-2 type, > say that a No-break space (NBSP) character is graphic. A graphic > character is defined as something that would use ink when being printed, > but a NBSP is just empty space with the special property that the words > on either side of it shouldn't be split across lines when being output > to, say, a display. > > Because many systems have similar defective locales, an error isn't > actually raised unless a threshold of failing tests is reached. This is > to prevent us from having to deal with failing smokes that aren't really > Perl bugs. But at some fairly small threshold of failure, it could be a > Perl bug that we should look at. What I decided rather arbitrarily, is > if 95% or more of the locales pass an individual test in this script, > then the failure is more likely to be in the failing locales than it is > to be in Perl, and so the test passes. But in your system, only 92% of > all the locales pass this one test. It turns out that it is a bug in > your locales, but the script did its job and notified us of a potential > problem. On other systems, I've ended up adjusting the threshold. On > AIX machines, almost all the locales have the same defect as your 8859-2 > ones do. The threshold there has been jacked up to if only 0.1% pass, > the script doesn't fail. > > (Note that your 8859-1 locales don't have the same problem, and yet they > have the NBSP character same as the 8859-2 ones. I wonder if the > definition of 8859-2 changed at some point, and these locale definitions > didn't get fully updated. Or maybe they all come from the same original > definition, and the person who wrote 8859-2 differed from the one who > wrote 8859-1, and wasn't as clued in, or they just were having a bad > day, or ...) > > Some of your other locales also are being found to have defects by this > test script, but in rates well below 5% of all the locales, so this is > the only failure that gets raised. (The other failures are shown if you > run the script by hand, but they are marked in such a way as to prevent > them from being shown as failures by the overall process.) > > I'll think about the appropriate way to change the test to not cause > this script to be marked as failing. > > But these are real defects that I believe should be reported to the OS > vendor. >
Karl, thanks for that explanation as to how lib/locale.t works. We -- P5P collectively -- need to come up with a Standard Operating Procedure for informing OSes of locale-related problems -- or, more generally, of problems which some OSes appear to have solved better than others. A smoke-tester who reports failures like these -- as Carlos Guevara has done at my request -- is not necessarily an expert in either locales or the operating system being reported on. So we as an open-source software project ought to figure out a way to communicate these problems with other such projects. Thank you very much. -- James E Keenan (jkeenan@cpan.org)
Subject: Re: [perl #131735] NetBSD 7.1 Smoke Test Failures
To: Karl Williamson <public [...] khwilliamson.com>
CC: perlbug-followup [...] perl.org
Date: Tue, 18 Jul 2017 22:37:17 -0500
From: Carlos Guevara <carlos [...] carlosguevara.com>
Download (untitled) / with headers
text/plain 797b
No problem.

Reports attached.


On Mon, Jul 17, 2017 at 4:40 PM, Karl Williamson <public@khwilliamson.com> wrote:
Show quoted text
On 07/17/2017 02:43 PM, Karl Williamson wrote:
Fixing that also correctly showed why netbsd is failing this test.  8% of the locales on your system, all of them being the ISO 8859-2 type, say that a No-break space (NBSP) character is graphic.  A graphic character is defined as something that would use ink when being printed, but a NBSP is just empty space with the special property that the words on either side of it shouldn't be split across lines when being output to, say, a display.


After a couple more minutes of thought (less than I thought would be necessary), I think I have a solution.  Please run the attached script, like before.

Download perl-Ilib-V.txt
text/plain 2.5k

Message body is not shown because sender requested not to inline it.

Message body not shown because it is not plain text.

RT-Send-CC: perl5-porters [...] perl.org
Download (untitled) / with headers
text/plain 2.5k
On Mon, 17 Jul 2017 20:46:07 GMT, public@khwilliamson.com wrote: Show quoted text
> [snip] > > Fixing that also correctly showed why netbsd is failing this test. 8% > of the locales on your system, all of them being the ISO 8859-2 type, > say that a No-break space (NBSP) character is graphic. A graphic > character is defined as something that would use ink when being printed, > but a NBSP is just empty space with the special property that the words > on either side of it shouldn't be split across lines when being output > to, say, a display. > > [snip]
Show quoted text
> What I decided rather arbitrarily, is > if 95% or more of the locales pass an individual test in this script, > then the failure is more likely to be in the failing locales than it is > to be in Perl, and so the test passes. But in your system, only 92% of > all the locales pass this one test. It turns out that it is a bug in > your locales, but the script did its job and notified us of a potential > problem. On other systems, I've ended up adjusting the threshold. On > AIX machines, almost all the locales have the same defect as your 8859-2 > ones do. The threshold there has been jacked up to if only 0.1% pass, > the script doesn't fail. > > (Note that your 8859-1 locales don't have the same problem, and yet they > have the NBSP character same as the 8859-2 ones. I wonder if the > definition of 8859-2 changed at some point, and these locale definitions > didn't get fully updated. Or maybe they all come from the same original > definition, and the person who wrote 8859-2 differed from the one who > wrote 8859-1, and wasn't as clued in, or they just were having a bad > day, or ...) > > [snip]
Show quoted text
> But these are real defects that I believe should be reported to the OS > vendor. >
This page describes how to submit a bug report to the NetBSD project: ##### http://www.netbsd.org/cgi-bin/sendpr.cgi?gndb=netbsd ##### I would be willing to be the submitter of such a bug report, but I would need: 1. From Carlos: a. The NetBSD release where you encountered the problem; b. Output of 'uname -a' on the smoke-testing rig. 2. From khw: a. Text for the "Full Description" box on that bug reporting page.* b. Text for the "How to repeat the problem" and "Fix to the problem if known" boxes on that page, if appropriate. *Karl, is there is some boilerplate text we could use in submitting such locale-related bug reports? If so, that would facilitate preparation of such reports in the future. So, if you all can supply the requested information, I will submit a report to NetBSD. Thank you very much. -- James E Keenan (jkeenan@cpan.org)
From: Carlos Guevara <carlos [...] carlosguevara.com>
Date: Sat, 22 Jul 2017 16:49:31 -0500
To: perlbug-followup [...] perl.org
Subject: Re: [perl #131735] NetBSD 7.1 Smoke Test Failures
Download (untitled) / with headers
text/plain 2.8k
On Thu, Jul 20, 2017 at 8:13 AM, James E Keenan via RT <perlbug-followup@perl.org> wrote:
Show quoted text
On Mon, 17 Jul 2017 20:46:07 GMT, public@khwilliamson.com wrote:
> [snip]
>
> Fixing that also correctly showed why netbsd is failing this test.  8%
> of the locales on your system, all of them being the ISO 8859-2 type,
> say that a No-break space (NBSP) character is graphic.  A graphic
> character is defined as something that would use ink when being printed,
> but a NBSP is just empty space with the special property that the words
> on either side of it shouldn't be split across lines when being output
> to, say, a display.
>
> [snip]

>  What I decided rather arbitrarily, is
> if 95% or more of the locales pass an individual test in this script,
> then the failure is more likely to be in the failing locales than it is
> to be in Perl, and so the test passes.  But in your system, only 92% of
> all the locales pass this one test.  It turns out that it is a bug in
> your locales, but the script did its job and notified us of a potential
> problem.  On other systems, I've ended up adjusting the threshold.  On
> AIX machines, almost all the locales have the same defect as your 8859-2
> ones do.  The threshold there has been jacked up to if only 0.1% pass,
> the script doesn't fail.
>
> (Note that your 8859-1 locales don't have the same problem, and yet they
> have the NBSP character same as the 8859-2 ones.  I wonder if the
> definition of 8859-2 changed at some point, and these locale definitions
> didn't get fully updated.  Or maybe they all come from the same original
> definition, and the person who wrote 8859-2 differed from the one who
> wrote 8859-1, and wasn't as clued in, or they just were having a bad
> day, or ...)
>
> [snip]

> But these are real defects that I believe should be reported to the OS
> vendor.
>

This page describes how to submit a bug report to the NetBSD project:

#####
http://www.netbsd.org/cgi-bin/sendpr.cgi?gndb=netbsd
#####

I would be willing to be the submitter of such a bug report, but I would need:

1. From Carlos:

a. The NetBSD release where you encountered the problem;

b. Output of 'uname -a' on the smoke-testing rig.

2. From khw:

a. Text for the "Full Description" box on that bug reporting page.*

b. Text for the "How to repeat the problem" and "Fix to the problem if known" boxes on that page, if appropriate.

*Karl, is there is some boilerplate text we could use in submitting such locale-related bug reports?  If so, that would facilitate preparation of such reports in the future.

So, if you all can supply the requested information, I will submit a report to NetBSD.

Thank you very much.

--
James E Keenan (jkeenan@cpan.org)

NetBSD Release:
NetBSD 7.1

uname -a:
NetBSD cjg-netbsd7.localdomain 7.1 NetBSD 7.1 (GENERIC.201703111743Z) amd64
Subject: Re: [perl #131735] NetBSD 7.1 Smoke Test Failures
To: Carlos Guevara <carlos [...] carlosguevara.com>
Date: Sat, 22 Jul 2017 22:59:30 -0600
From: Karl Williamson <public [...] khwilliamson.com>
CC: perlbug-followup [...] perl.org
On 07/18/2017 09:37 PM, Carlos Guevara wrote: Show quoted text
> No problem. > > Reports attached.
That didn't fix it. I think it's a casing problem, where I was expecting NetBSD to be in all lowercase. Try the attached, please. Show quoted text
> > > On Mon, Jul 17, 2017 at 4:40 PM, Karl Williamson > <public@khwilliamson.com <mailto:public@khwilliamson.com>> wrote: > > On 07/17/2017 02:43 PM, Karl Williamson wrote: > > Fixing that also correctly showed why netbsd is failing this > test. 8% of the locales on your system, all of them being the > ISO 8859-2 type, say that a No-break space (NBSP) character is > graphic. A graphic character is defined as something that would > use ink when being printed, but a NBSP is just empty space with > the special property that the words on either side of it > shouldn't be split across lines when being output to, say, a > display. > > > > After a couple more minutes of thought (less than I thought would be > necessary), I think I have a solution. Please run the attached > script, like before. > >
Download locale.t
text/plain 95.4k

Message body is not shown because sender requested not to inline it.

Subject: Re: [perl #131735] NetBSD 7.1 Smoke Test Failures
To: perlbug-followup [...] perl.org
Date: Sun, 23 Jul 2017 22:41:42 -0500
From: Carlos Guevara <carlos [...] carlosguevara.com>
Download (untitled) / with headers
text/plain 1.2k


On Sun, Jul 23, 2017 at 12:00 AM, karl williamson via RT <perlbug-followup@perl.org> wrote:
Show quoted text
On 07/18/2017 09:37 PM, Carlos Guevara wrote:
> No problem.
>
> Reports attached.

That didn't fix it.  I think it's a casing problem, where I was
expecting NetBSD to be in all lowercase.  Try the attached, please.
>
>
> On Mon, Jul 17, 2017 at 4:40 PM, Karl Williamson
> <public@khwilliamson.com <mailto:public@khwilliamson.com>> wrote:
>
>     On 07/17/2017 02:43 PM, Karl Williamson wrote:
>
>         Fixing that also correctly showed why netbsd is failing this
>         test.  8% of the locales on your system, all of them being the
>         ISO 8859-2 type, say that a No-break space (NBSP) character is
>         graphic.  A graphic character is defined as something that would
>         use ink when being printed, but a NBSP is just empty space with
>         the special property that the words on either side of it
>         shouldn't be split across lines when being output to, say, a
>         display.
>
>
>
>     After a couple more minutes of thought (less than I thought would be
>     necessary), I think I have a solution.  Please run the attached
>     script, like before.
>
>


No problem.

Reports attached.

Download perl-Ilib-V.txt
text/plain 2.5k

Message body is not shown because sender requested not to inline it.

Message body not shown because it is not plain text.

Subject: Re: [perl #131735] NetBSD 7.1 Smoke Test Failures
From: Karl Williamson <public [...] khwilliamson.com>
Date: Sun, 30 Jul 2017 22:41:19 -0600
To: Carlos Guevara <carlos [...] carlosguevara.com>, perlbug-followup [...] perl.org
Download (untitled) / with headers
text/plain 1.7k
On 07/23/2017 09:41 PM, Carlos Guevara wrote: Show quoted text
> > > On Sun, Jul 23, 2017 at 12:00 AM, karl williamson via RT > <perlbug-followup@perl.org <mailto:perlbug-followup@perl.org>> wrote: > > On 07/18/2017 09:37 PM, Carlos Guevara wrote:
> > No problem. > > > > Reports attached.
> > That didn't fix it. I think it's a casing problem, where I was > expecting NetBSD to be in all lowercase. Try the attached, please.
> > > > > > On Mon, Jul 17, 2017 at 4:40 PM, Karl Williamson > > <public@khwilliamson.com <mailto:public@khwilliamson.com>
> <mailto:public@khwilliamson.com <mailto:public@khwilliamson.com>>> > wrote:
> > > > On 07/17/2017 02:43 PM, Karl Williamson wrote: > > > > Fixing that also correctly showed why netbsd is failing this > > test. 8% of the locales on your system, all of them
> being the
> > ISO 8859-2 type, say that a No-break space (NBSP)
> character is
> > graphic. A graphic character is defined as something
> that would
> > use ink when being printed, but a NBSP is just empty
> space with
> > the special property that the words on either side of it > > shouldn't be split across lines when being output to, say, a > > display. > > > > > > > > After a couple more minutes of thought (less than I thought
> would be
> > necessary), I think I have a solution. Please run the attached > > script, like before. > > > >
> > > No problem. > > Reports attached. >
When I finally got around to looking at why my file didn't fix the problem, I realized I had misspelled a locale name. A new test file is attached.
Download locale.t
text/plain 95.4k

Message body is not shown because sender requested not to inline it.

Date: Mon, 31 Jul 2017 00:43:46 -0500
To: perlbug-followup [...] perl.org
From: Carlos Guevara <carlos [...] carlosguevara.com>
Subject: Re: [perl #131735] NetBSD 7.1 Smoke Test Failures
Download (untitled) / with headers
text/plain 1.9k
On Sun, Jul 30, 2017 at 11:41 PM, karl williamson via RT <perlbug-followup@perl.org> wrote:
Show quoted text
On 07/23/2017 09:41 PM, Carlos Guevara wrote:
>
>
> On Sun, Jul 23, 2017 at 12:00 AM, karl williamson via RT
> <perlbug-followup@perl.org <mailto:perlbug-followup@perl.org>> wrote:
>
>     On 07/18/2017 09:37 PM, Carlos Guevara wrote:
>      > No problem.
>      >
>      > Reports attached.
>
>     That didn't fix it.  I think it's a casing problem, where I was
>     expecting NetBSD to be in all lowercase.  Try the attached, please.
>      >
>      >
>      > On Mon, Jul 17, 2017 at 4:40 PM, Karl Williamson
>      > <public@khwilliamson.com <mailto:public@khwilliamson.com>
>     <mailto:public@khwilliamson.com <mailto:public@khwilliamson.com>>>
>     wrote:
>      >
>      >     On 07/17/2017 02:43 PM, Karl Williamson wrote:
>      >
>      >         Fixing that also correctly showed why netbsd is failing this
>      >         test.  8% of the locales on your system, all of them
>     being the
>      >         ISO 8859-2 type, say that a No-break space (NBSP)
>     character is
>      >         graphic.  A graphic character is defined as something
>     that would
>      >         use ink when being printed, but a NBSP is just empty
>     space with
>      >         the special property that the words on either side of it
>      >         shouldn't be split across lines when being output to, say, a
>      >         display.
>      >
>      >
>      >
>      >     After a couple more minutes of thought (less than I thought
>     would be
>      >     necessary), I think I have a solution.  Please run the attached
>      >     script, like before.
>      >
>      >
>
>
> No problem.
>
> Reports attached.
>

When I finally got around to looking at why my file didn't fix the
problem, I realized I had misspelled a locale name.  A new test file is
attached.

No problem.

Reports attached.

Download perl-Ilib-V.txt
text/plain 2.5k

Message body is not shown because sender requested not to inline it.

Message body not shown because it is not plain text.

Date: Mon, 7 Aug 2017 11:44:38 -0600
Subject: Re: [perl #131735] NetBSD 7.1 Smoke Test Failures
From: Karl Williamson <public [...] khwilliamson.com>
To: Carlos Guevara <carlos [...] carlosguevara.com>, perlbug-followup [...] perl.org
Download (untitled) / with headers
text/plain 3.6k
On 07/22/2017 03:49 PM, Carlos Guevara wrote: Show quoted text
> On Thu, Jul 20, 2017 at 8:13 AM, James E Keenan via RT > <perlbug-followup@perl.org <mailto:perlbug-followup@perl.org>> wrote:
The changes to locale.t are now in blead. This should fix this portion of the ticket, but I'll await confirmation. I don't know what to do about the other portions. Jim? Show quoted text
> > On Mon, 17 Jul 2017 20:46:07 GMT, public@khwilliamson.com > <mailto:public@khwilliamson.com> wrote:
> > [snip] > > > > Fixing that also correctly showed why netbsd is failing this
> test. 8%
> > of the locales on your system, all of them being the ISO 8859-2 type, > > say that a No-break space (NBSP) character is graphic. A graphic > > character is defined as something that would use ink when being
> printed,
> > but a NBSP is just empty space with the special property that the
> words
> > on either side of it shouldn't be split across lines when being
> output
> > to, say, a display. > > > > [snip]
>
> > What I decided rather arbitrarily, is > > if 95% or more of the locales pass an individual test in this script, > > then the failure is more likely to be in the failing locales than
> it is
> > to be in Perl, and so the test passes. But in your system, only
> 92% of
> > all the locales pass this one test. It turns out that it is a bug in > > your locales, but the script did its job and notified us of a
> potential
> > problem. On other systems, I've ended up adjusting the
> threshold. On
> > AIX machines, almost all the locales have the same defect as your
> 8859-2
> > ones do. The threshold there has been jacked up to if only 0.1%
> pass,
> > the script doesn't fail. > > > > (Note that your 8859-1 locales don't have the same problem, and
> yet they
> > have the NBSP character same as the 8859-2 ones. I wonder if the > > definition of 8859-2 changed at some point, and these locale
> definitions
> > didn't get fully updated. Or maybe they all come from the same
> original
> > definition, and the person who wrote 8859-2 differed from the one who > > wrote 8859-1, and wasn't as clued in, or they just were having a bad > > day, or ...) > > > > [snip]
>
> > But these are real defects that I believe should be reported to
> the OS
> > vendor. > >
> > This page describes how to submit a bug report to the NetBSD project: > > ##### > http://www.netbsd.org/cgi-bin/sendpr.cgi?gndb=netbsd > <http://www.netbsd.org/cgi-bin/sendpr.cgi?gndb=netbsd> > ##### > > I would be willing to be the submitter of such a bug report, but I > would need: > > 1. From Carlos: > > a. The NetBSD release where you encountered the problem; > > b. Output of 'uname -a' on the smoke-testing rig. > > 2. From khw: > > a. Text for the "Full Description" box on that bug reporting page.* > > b. Text for the "How to repeat the problem" and "Fix to the problem > if known" boxes on that page, if appropriate. > > *Karl, is there is some boilerplate text we could use in submitting > such locale-related bug reports? If so, that would facilitate > preparation of such reports in the future. > > So, if you all can supply the requested information, I will submit a > report to NetBSD. > > Thank you very much. > > -- > James E Keenan (jkeenan@cpan.org <mailto:jkeenan@cpan.org>) > > > NetBSD Release: > NetBSD 7.1 > > uname -a: > NetBSD cjg-netbsd7.localdomain 7.1 NetBSD 7.1 (GENERIC.201703111743Z) amd64
RT-Send-CC: perl5-porters [...] perl.org
Download (untitled) / with headers
text/plain 1.6k
On Mon, 07 Aug 2017 17:45:42 GMT, public@khwilliamson.com wrote: Show quoted text
> On 07/22/2017 03:49 PM, Carlos Guevara wrote:
> > On Thu, Jul 20, 2017 at 8:13 AM, James E Keenan via RT > > <perlbug-followup@perl.org <mailto:perlbug-followup@perl.org>> wrote:
> > The changes to locale.t are now in blead. This should fix this > portion > of the ticket, but I'll await confirmation. I don't know what to do > about the other portions. Jim?
> > > > > > > > NetBSD Release: > > NetBSD 7.1 > > > > uname -a: > > NetBSD cjg-netbsd7.localdomain 7.1 NetBSD 7.1 (GENERIC.201703111743Z) > > amd64
The only other smoke testing failure being reported on this platform (http://perl.develop-help.com/raw/?id=200535) looks like this: ##### Failures: (common-args) none [perlio] -DDEBUGGING -Duseithreads Inconsistent test results (between TEST and harness): ../cpan/Sys-Syslog/t/syslog.t.... FAILED at test 147 ##### I confess that I never quite know how to react to cases where t/TEST and t/harness come up with different results. I only smoke with t/harness because the VMs in which I smoke are situated on my everyday laptop. And I only have the energy to focus on tests which are failing very consistently. That's even more true on platforms like this one where we're lucky to have one or two people submitting smoke reports -- much less someone who can actually diagnose and correct problems. So, I'm going to mark this ticket Resolved because we (P5P contributors) have accomplished as much as we reasonably can. (Send me some appropriate language and I'll open up a bug ticket with NetBSD about the locales.) Thank you very much. -- James E Keenan (jkeenan@cpan.org)
From: Dave Mitchell <davem [...] iabyn.com>
Subject: Re: [perl #131735] NetBSD 7.1 Smoke Test Failures
CC: perl5-porters [...] perl.org
Date: Tue, 8 Aug 2017 08:21:54 +0100
To: James E Keenan via RT <perlbug-followup [...] perl.org>
Download (untitled) / with headers
text/plain 652b
On Mon, Aug 07, 2017 at 07:27:32PM -0700, James E Keenan via RT wrote: Show quoted text
> I confess that I never quite know how to react to cases where t/TEST and > t/harness come up with different results.
There are two common reasons for a difference. First the failure may be intermittent, and whether it failed under TEST or harness is just random (e.g. random hash seed, race condition on parallel test etc). Secondly, TEST sets PERL_DESTRUCT_LEVEL=2 (which harness doesn't) which tends to show up more problems. -- Technology is dominated by two types of people: those who understand what they do not manage, and those who manage what they do not understand.


This service is sponsored and maintained by Best Practical Solutions and runs on Perl.org infrastructure.

For issues related to this RT instance (aka "perlbug"), please contact perlbug-admin at perl.org