f



Problem with "make check" in building GAWK 4.1.4

Building GAWK 4.1.4 on a Raspberry Pi, with GMP/MPFR.
Have done this many times before - generally everything is OK, except...

configure/make routine OK for all 3 packages (MPFR 3.1.4, GMP 6.1.1), make
check works fine for GMP and MPFR, but fails on GAWK itself, as shown below
(this is the tail end of the "make check" output, cut-and-pasted from the
screen) :

--- Cut Here ---
make[2]: Entering directory '/home/pi/Build/gawk-4.1.4/test'
2 TESTS FAILED
Makefile:4128: recipe for target 'pass-fail' failed
make[2]: *** [pass-fail] Error 1
make[2]: Leaving directory '/home/pi/Build/gawk-4.1.4/test'
make[2]: Entering directory '/home/pi/Build/gawk-4.1.4/test'
for i in _* ; \
do  \
	if [ "$i" != "_*" ]; then \
	echo ============== $i ============= ; \
	base=`echo $i | sed 's/^_//'` ; \
	if [ -r ${base}.ok ]; then \
	diff -c ${base}.ok $i ; \
	else \
	diff -c "."/${base}.ok  $i ; \
	fi ; \
	fi ; \
done | more
============== _mbstr1 =============
*** mbstr1.ok	Mon Jan 25 11:49:20 2016
--- _mbstr1	Sat Sep 10 12:22:28 2016
***************
*** 1,3 ****
- gawk: mbstr1.awk:2: warning: Invalid multibyte data detected. There may be a m
ismatch between your data and your locale.
  4
  1
--- 1,2 ----
============== _mbstr2 =============
*** mbstr2.ok	Mon Jan 25 11:49:20 2016
--- _mbstr2	Sat Sep 10 12:22:28 2016
***************
*** 1,5 ****
  2007
- gawk: mbstr2.awk:1: (FILENAME=- FNR=2) warning: Invalid multibyte data detecte
d. There may be a mismatch between your data and your locale.
  2007
  2007
  2007
--- 1,4 ----
make[2]: Leaving directory '/home/pi/Build/gawk-4.1.4/test'
Makefile:1651: recipe for target 'check' failed
make[1]: *** [check] Error 1
make[1]: Leaving directory '/home/pi/Build/gawk-4.1.4/test'
Makefile:743: recipe for target 'check-recursive' failed
make: *** [check-recursive] Error 1
--- Cut Here ---

Everything else about the binary seems to work fine, and I'm inclined to
just ignore this error.  I note that the 'mb' testing is the very last part
of the test suite, so, again, I'm inclined to just live with it.

Still, I'm curious as to why it fails.  Obviously, this is for me to debug,
and I'm certainly willing and able to do so, but I'd like some help as to
what exactly the above errors mean.  What is actually testing/comparing?
What environment variables does it depend on?

If somebody would give me some help on what exactly is being compared (and
why), I'll probably be able to fix it.  Also, some help on what it really
means - I.e., if/why I should even care about it.

-- 
The randomly chosen signature file that would have appeared here is more than 4
lines long.  As such, it violates one or more Usenet RFCs.  In order to remain in
compliance with said RFCs, the actual sig can be found at the following web address:
	http://www.xmission.com/~gazelle/Sigs/ForFoxViewers
0
gazelle
9/10/2016 4:40:26 PM
comp.lang.awk 3450 articles. 0 followers. Post Follow

11 Replies
588 Views

Similar Articles

[PageSpeed] 55

On 10.09.2016 18:40, Kenny McCormack wrote:
> Building GAWK 4.1.4 on a Raspberry Pi, with GMP/MPFR.
> Have done this many times before - generally everything is OK, except...
> 
> configure/make routine OK for all 3 packages (MPFR 3.1.4, GMP 6.1.1), make
> check works fine for GMP and MPFR, but fails on GAWK itself, as shown below
> (this is the tail end of the "make check" output, cut-and-pasted from the
> screen) :
> 
> --- Cut Here ---
> make[2]: Entering directory '/home/pi/Build/gawk-4.1.4/test'
> 2 TESTS FAILED
> Makefile:4128: recipe for target 'pass-fail' failed
> make[2]: *** [pass-fail] Error 1
> make[2]: Leaving directory '/home/pi/Build/gawk-4.1.4/test'
> make[2]: Entering directory '/home/pi/Build/gawk-4.1.4/test'
> for i in _* ; \
> do  \
> 	if [ "$i" != "_*" ]; then \
> 	echo ============== $i ============= ; \
> 	base=`echo $i | sed 's/^_//'` ; \
> 	if [ -r ${base}.ok ]; then \
> 	diff -c ${base}.ok $i ; \
> 	else \
> 	diff -c "."/${base}.ok  $i ; \
> 	fi ; \
> 	fi ; \
> done | more
> ============== _mbstr1 =============
> *** mbstr1.ok	Mon Jan 25 11:49:20 2016
> --- _mbstr1	Sat Sep 10 12:22:28 2016
> ***************
> *** 1,3 ****
> - gawk: mbstr1.awk:2: warning: Invalid multibyte data detected. There may be a m
> ismatch between your data and your locale.
>   4
>   1
> --- 1,2 ----
> ============== _mbstr2 =============
> *** mbstr2.ok	Mon Jan 25 11:49:20 2016
> --- _mbstr2	Sat Sep 10 12:22:28 2016
> ***************
> *** 1,5 ****
>   2007
> - gawk: mbstr2.awk:1: (FILENAME=- FNR=2) warning: Invalid multibyte data detecte
> d. There may be a mismatch between your data and your locale.
>   2007
>   2007
>   2007
> --- 1,4 ----
> make[2]: Leaving directory '/home/pi/Build/gawk-4.1.4/test'
> Makefile:1651: recipe for target 'check' failed
> make[1]: *** [check] Error 1
> make[1]: Leaving directory '/home/pi/Build/gawk-4.1.4/test'
> Makefile:743: recipe for target 'check-recursive' failed
> make: *** [check-recursive] Error 1
> --- Cut Here ---
> 
> Everything else about the binary seems to work fine, and I'm inclined to
> just ignore this error.  I note that the 'mb' testing is the very last part
> of the test suite, so, again, I'm inclined to just live with it.
> 
> Still, I'm curious as to why it fails.  Obviously, this is for me to debug,
> and I'm certainly willing and able to do so, but I'd like some help as to
> what exactly the above errors mean.  What is actually testing/comparing?
> What environment variables does it depend on?
> 
> If somebody would give me some help on what exactly is being compared (and
> why), I'll probably be able to fix it.  Also, some help on what it really
> means - I.e., if/why I should even care about it.

After trying make check on my system, and inspecting the test cases, just a
few comments...

First the test suite informs:

  Locale environment:
	LC_ALL="C" LANG="C"

  ======== Starting basic tests ========

so I'd have _expected_ it to test in a correct predefined environment. (But
that may also be just misleading.)

My own make check returns only a warning:

  ======== Done with machine-specific tests ========
  %%%%%%%%%% Inadequate locale support: skipping charset tests.
  %%%%%%%%%% At least en_US.UTF-8, ru_RU.UTF-8 and ja_JP.UTF-8 are needed.
  ======== Starting shared library tests ========

I'm running mixed locales; LC_* are all defined as de_DE.UTF-8, LANG* as
en_US.utf8, but generally using an UTF-8 encoding.

Then the errors appear in the testcases: ~/test/mbstr[01].*
It's purpose is to locate unexpected character encodings in data provided
to various string functions; the messages you see come from those test cases
triggered. At least one of the two testcases stems from an UTF-8 envoronment
(and I suppose there shouldn't be an issues in any 8-bit character encoding
envonronment).

Janis

0
Janis
9/10/2016 5:23:31 PM
In article <nr1d1q$dn5$1@news.xmission.com>,
Kenny McCormack <gazelle@shell.xmission.com> wrote:
>Building GAWK 4.1.4 on a Raspberry Pi, with GMP/MPFR.
>Have done this many times before - generally everything is OK, except...
>
>configure/make routine OK for all 3 packages (MPFR 3.1.4, GMP 6.1.1), make
>check works fine for GMP and MPFR, but fails on GAWK itself, as shown below
>(this is the tail end of the "make check" output, cut-and-pasted from the
>screen) :

I should add, just to be clear, that I have built it (both the current and
earlier versions) on both i386 and x64 and everything is fine (including
the 'make check').  So, it seems to me that it is something either unique
to the Rpi (i.e., most likely, the software setup of the Raspbian OS) or
something about my particular configuration there (not that there is
anything particularly strange about it).

But as Janis points out, it should be environment-agnostic - it should be
(and it seems it is) setting its own environment vairables, so that it
should "just work".

One difference is that the Pi is multi-core (I don't think any of my
i386/x64 machines are), and so I've dabbled in using "make -j 4" to speed
things up (might as well get all those cores working, right?).  I suspect
that this may, in fact, be some part of the problem.  But it is tedious to
do all the needed experiments to be sure.  And note that the last time I
tested this (today), I did *not* use "-j 4" on the GAWK invocation of make.
(I *may* have used it in building either GMP and/or MPFR, but the details
are somewhat lost in the mists of time...)

And if it is related to multi-core compilation, why would it affect (only)
the "multi-byte" stuff?  I'd like to hear from any of the developers if
there is or might be any issue with using multi-core compilation.

-- 
Mike Huckabee has yet to consciously uncouple from Josh Duggar.

0
gazelle
9/10/2016 5:35:28 PM
In article <nr1g90$fia$1@news.xmission.com>,
Kenny McCormack <gazelle@shell.xmission.com> wrote:
>I'd like to hear from any of the developers if
>there is or might be any issue with using multi-core compilation.

There is generally no issue with multicore compilation. I use make -j all
the time.

The failing tests indicate some kind of inadequate locale support
on your system. You can try to figure out how to install more
locales or just ignore those tests.
-- 
Aharon (Arnold) Robbins 		arnold AT skeeve DOT com
0
arnold
9/10/2016 7:28:53 PM
In article <nr1mtl$m74$1@dont-email.me>,
Aharon Robbins <arnold@skeeve.com> wrote:
>In article <nr1g90$fia$1@news.xmission.com>,
>Kenny McCormack <gazelle@shell.xmission.com> wrote:
>>I'd like to hear from any of the developers if
>>there is or might be any issue with using multi-core compilation.
>
>There is generally no issue with multicore compilation. I use make -j all
>the time.

OK.  Good to know.
BTW, do you have any knowledge about specifically compiling on the RPi?
(I.e., what people have done - what works - what doesn't)

Note that it sometimes works.  I won't say it is random - it is probably
dependent on something - but, just as clearly, I don't know what that
something is.  And, as I said, it would be too much work/time for me to
design and execute a full test-suite to determine what causes it.

>The failing tests indicate some kind of inadequate locale support
>on your system. You can try to figure out how to install more
>locales or just ignore those tests.

Well, when you think about it, that does sound a little odd.

Why should a misconfiguration of the host system (presumably, on the
part of the Raspbian packagers/maintainers) result in a message (and thus
the perception) to the effect that something went wrong in the compilation
of GAWK?

Anyway, is there any way to turn off those tests (i.e. to tell it to skip
them)?  I.e., so that you could get a clean "All Tests Passed" kind of
message at the end (because, for example, maybe you need to show that to
your boss or maybe it is required by corporate standards or something like
that) ?

-- 
The randomly chosen signature file that would have appeared here is more than 4
lines long.  As such, it violates one or more Usenet RFCs.  In order to remain in
compliance with said RFCs, the actual sig can be found at the following web address:
	http://www.xmission.com/~gazelle/Sigs/Seneca
0
gazelle
9/10/2016 8:55:48 PM
In article <nr1s0k$nko$1@news.xmission.com>,
>BTW, do you have any knowledge about specifically compiling on the RPi?
>(I.e., what people have done - what works - what doesn't)

None at all.

>>The failing tests indicate some kind of inadequate locale support
>>on your system. You can try to figure out how to install more
>>locales or just ignore those tests.
>
>Well, when you think about it, that does sound a little odd.

Or else it's libc weirdness with respect to handling multibyte data.

>Anyway, is there any way to turn off those tests (i.e. to tell it to skip
>them)?

Edit the Makefile after configure runs and before running 'make check'.
-- 
Aharon (Arnold) Robbins 		arnold AT skeeve DOT com
0
arnold
9/11/2016 2:43:29 AM
In article <nr2gch$f7k$1@dont-email.me>,
none) (Aharon Robbins <arnold@tanda.> wrote:
>In article <nr1s0k$nko$1@news.xmission.com>,
>>BTW, do you have any knowledge about specifically compiling on the RPi?
>>(I.e., what people have done - what works - what doesn't)
>
>None at all.

Understandable.  Just as an aside, you might want to get one.  They are
pretty cool little boards.

But in any case, the hope is that someone on the RPi side - maybe the
Raspbian maintainers - might see this thread and get around to fixing
things.

>>>The failing tests indicate some kind of inadequate locale support
>>>on your system. You can try to figure out how to install more
>>>locales or just ignore those tests.
>>
>>Well, when you think about it, that does sound a little odd.
>
>Or else it's libc weirdness with respect to handling multibyte data.

Right.  But that just reinforces the point.  Which is that whenever you're
testing anything, you're always comparing two things - One that is assumed
to be good and the other is the thing you are testing.  In this case, it
seems like we're assuming that the GAWK program (the thing you just
compiled) is good and that the thing we're testing is the OS (i.e., the OS
configuration).  Interesting, that...

>>Anyway, is there any way to turn off those tests (i.e. to tell it to skip
>>them)?
>
>Edit the Makefile after configure runs and before running 'make check'.

OK.  I'll try that.

>Aharon (Arnold) Robbins 		arnold AT skeeve DOT com

BTW, because of the way your email address is setup on this (and only
this) post, your name shows up in my newsreader (in the article list display)
as "(none)", rather than as "Aharon Robbins" (as it does in every other post
from you).

-- 

    "This ain't my first time at the rodeo"

is a line from the movie, Mommie Dearest, said by Joan Crawford at a board meeting.
0
gazelle
9/11/2016 1:03:35 PM
On 11.09.2016 15:03, Kenny McCormack wrote:
> In article <nr2gch$f7k$1@dont-email.me>,
> none) (Aharon Robbins <arnold@tanda.> wrote:
>> [...]
> 
> BTW, because of the way your email address is setup on this (and only
> this) post, your name shows up in my newsreader (in the article list display)
> as "(none)", rather than as "Aharon Robbins" (as it does in every other post
> from you).

Depends in the newsreader how it parses his new address in the header:
arnold@tanda.(none) (Aharon Robbins)
Obviously your trn misinterprets the first parenthesis as name part.
With Thunderbird I still see his name.
(I can't comment on whether "arnold@tanda.(none)" is standard or nor.)

Janis

0
Janis
9/11/2016 1:42:19 PM
In article <nr3kn7$svs$2@news.xmission.com>,
Kenny McCormack <gazelle@shell.xmission.com> wrote:
>Understandable.  Just as an aside, you might want to get one.  They are
>pretty cool little boards.

Nice of you to spend my money.

But do feel free to send me one if it's really important to you.
-- 
Aharon (Arnold) Robbins 		arnold AT skeeve DOT com
0
arnold
9/11/2016 1:49:09 PM
In article <nr3nck$pl3$1@dont-email.me>,
Aharon Robbins <arnold@skeeve.com> wrote:
>In article <nr3kn7$svs$2@news.xmission.com>,
>Kenny McCormack <gazelle@shell.xmission.com> wrote:
>>Understandable.  Just as an aside, you might want to get one.  They are
>>pretty cool little boards.
>
>Nice of you to spend my money.
>
>But do feel free to send me one if it's really important to you.

One cannot make a friendly suggestion here without getting crapped on?

-- 
They say compassion is a virtue, but I don't have the time!

    - David Byrne -

0
gazelle
9/11/2016 1:58:34 PM
In article <nr3nua$svs$3@news.xmission.com>,
Kenny McCormack <gazelle@shell.xmission.com> wrote:
>>Nice of you to spend my money.
>>
>>But do feel free to send me one if it's really important to you.
>
>One cannot make a friendly suggestion here without getting crapped on?

You have admitted that you are not interested in investigating
the rPi issues w.r.t. gawk.

If you already have one but don't care enough about it, why should I
spend my time or money on an rPi?  I spend way too much of my time on gawk
as it is now.

Thanks, but no thanks.
-- 
Aharon (Arnold) Robbins 		arnold AT skeeve DOT com
0
arnold
9/11/2016 5:53:05 PM
In article <nr45m1$c6r$1@dont-email.me>,
Aharon Robbins <arnold@skeeve.com> wrote:
>In article <nr3nua$svs$3@news.xmission.com>,
>Kenny McCormack <gazelle@shell.xmission.com> wrote:
>>>Nice of you to spend my money.
>>>
>>>But do feel free to send me one if it's really important to you.
>>
>>One cannot make a friendly suggestion here without getting crapped on?
>
>You have admitted that you are not interested in investigating
>the rPi issues w.r.t. gawk.
>
>If you already have one but don't care enough about it, why should I
>spend my time or money on an rPi?  I spend way too much of my time on gawk
>as it is now.

There's really no need for this vitriol.  I get it now, you misunderstood
my suggestion as being some sort of one or both of the following:

    1) You should do this to satisfy me.
    2) You're a lazy (and cheap!) slug for not doing so.

Needless to say, neither of the above apply.  I was merely suggesting that
if you have $35 to spare, you might find it fun to get one and play with it.
It obviously doesn't matter one way or the other to me what you do.

But it is business as usual on Usenet, to always assume the worst about
people's motives.

P.S.  You can exit this mess gracefully by just not responding further on
this sub-thread.

-- 
The randomly chosen signature file that would have appeared here is more than 4
lines long.  As such, it violates one or more Usenet RFCs.  In order to remain in
compliance with said RFCs, the actual sig can be found at the following web address:
	http://www.xmission.com/~gazelle/Sigs/FreeCollege
0
gazelle
9/11/2016 10:53:21 PM
Reply: