[fricas-devel] )trace )stats reset -- abort s xx.input?

[Raymond Rogers]

I have an input file that I want to trace, a subroutine with different
arguments.
After the first test argument and call I do
)trace )stats
)trace )stats reset
and it stops without going on to the second test argument?
Is this expected?
The recording of the first argument times, counts, arguments; is as
expected.  But I would like to carry on into the next.  I do have
spooling on during this.

Code section:
)system rm trace_rec_factor_poly.log
)spool trace_rec_factor_poly.log
p1:XDP :=(3*x*y+5*y*x)^3
)trace rec_factor_poly  )only a )time )count   )math
p1r := rec_factor_poly(p1)
)trace )stats
)trace )stats reset
-
)trace rec_factor_poly  )only a )time )count   )math
p2:XDP :=(3*x*y+5*y*x+y*z)^3
p2r := rec_factor_poly(p2)
)trace )stats
)trace )stats reset
-
)spool

Ray

[oldk1331]

Hi, can you make a simpler reproducible example?
I'm not sure about what you are describing.

[Raymond Rogers]

Hi, Since there is a certain flakiness that occurs in my system.  I
decided to check the installation.  The compile configuration picked up
   The function BOOT::/UNTRACE-2 is undefined.

I recompiled and have the make file output, and, in fact, it has
UNTRACE-2 as undefined.  Does this indicate I have a bad lisp or some such?

I have attached some test files that are also mysterious to me. But
minor changes change the errors and such so a large part is probably my
ignorance; sort of like quantum mechanics, the bugs are dead and alive.
Misc
Startup:
viewman not present, disabling graphics
/bin/sh: 1: exec:
/usr/local/lib/fricas/target/x86_64-linux-gnu/bin/hypertex: not found
Checking for foreign routines
AXIOM="/usr/local/lib/fricas/target/x86_64-linux-gnu"
spad-lib="/usr/local/lib/fricas/target/x86_64-linux-gnu/lib/libspad.so"
foreign routines found
openServer result 0
                       FriCAS Computer Algebra System
                            Version: FriCAS 1.3.4
                   Timestamp: Sat Jul 14 18:19:47 EDT 2018

System:
Ubuntu
Release 16.04.4 LTS (Xenial Xerus) 64-bit
Kernel Linux 4.4.0-130-generic x86_64
MATE 1.12.1

Oh yes, I am more that willing to change Lisp's or admit to errors. 
Although being constructive is appreciated.

Ray

Post by oldk1331
Hi, can you make a simpler reproducible example?
I'm not sure about what you are describing.

[oldk1331]

On Sun, Jul 15, 2018 at 10:58 AM, Raymond Rogers

Hi, Since there is a certain flakiness that occurs in my system. I decided
to check the installation. The compile configuration picked up sbcl 1.4.5;
The function BOOT::/UNTRACE-2 is undefined.
I recompiled and have the make file output, and, in fact, it has UNTRACE-2
as undefined. Does this indicate I have a bad lisp or some such?

It seems this function is indeed undefined, so not a lisp problem.
Can you describe how to trigger this error?

I have attached some test files that are also mysterious to me. But minor
changes change the errors and such so a large part is probably my ignorance;
sort of like quantum mechanics, the bugs are dead and alive.

In your attached file, there is an error in
facc(n) == if n < 3 then n else n * fac (n - 1)
it should be "facc" instead of "fac", is that intentional?

Misc
viewman not present, disabling graphics
/usr/local/lib/fricas/target/x86_64-linux-gnu/bin/hypertex: not found
Checking for foreign routines
AXIOM="/usr/local/lib/fricas/target/x86_64-linux-gnu"
spad-lib="/usr/local/lib/fricas/target/x86_64-linux-gnu/lib/libspad.so"
foreign routines found
openServer result 0
FriCAS Computer Algebra System
Version: FriCAS 1.3.4
Timestamp: Sat Jul 14 18:19:47 EDT 2018
Ubuntu
Release 16.04.4 LTS (Xenial Xerus) 64-bit
Kernel Linux 4.4.0-130-generic x86_64
MATE 1.12.1
Oh yes, I am more that willing to change Lisp's or admit to errors.
Although being constructive is appreciated.
Ray

[Raymond Rogers]

Comments with leading ----

Post by oldk1331
On Sun, Jul 15, 2018 at 10:58 AM, Raymond Rogers

Hi, Since there is a certain flakiness that occurs in my system. I decided
to check the installation. The compile configuration picked up sbcl 1.4.5;
The function BOOT::/UNTRACE-2 is undefined.
I recompiled and have the make file output, and, in fact, it has UNTRACE-2
as undefined. Does this indicate I have a bad lisp or some such?

It seems this function is indeed undefined, so not a lisp problem.
Can you describe how to trigger this error?

--- When it occurs again I will preserve it, instead of trying to fix
it, and send it.  The problem is, it typically occurs inside of larger
files.  Since it's not intended I will curb my frustration.  I will try
to shrink it though; it's worth getting rid of bogus calls (or whatever)
.  There are a bunch more uncompiled functions in my make.log,  is the
list of any interest?  I could grep inside of the source files if that
is of any use; if it's not compiled it shouldn't be evoked/called.
Unless it's one of those compiled on demand cases.   Although I am not
familiar enough with spad/boot/lisp to be good at digging in I can make
a concordance (I think).  In the past I have found concordance's usefull.

Post by oldk1331

I have attached some test files that are also mysterious to me. But minor
changes change the errors and such so a large part is probably my ignorance;
sort of like quantum mechanics, the bugs are dead and alive.

In your attached file, there is an error in
facc(n) == if n < 3 then n else n * fac (n - 1)
it should be "facc" instead of "fac", is that intentional?

--- Sigh (:  (:  I triple checked the wrong places, will fix. :-! I will
get back to )trace )reset -> abort (without any error) now; as you can
see I was trying for that with my sent code.  I do see my
troubleshooting error; I should have done an objective (i.e. editor)
scan instead of a visual (biased) scan :-(

Post by oldk1331

Misc
viewman not present, disabling graphics
/usr/local/lib/fricas/target/x86_64-linux-gnu/bin/hypertex: not found
Checking for foreign routines
AXIOM="/usr/local/lib/fricas/target/x86_64-linux-gnu"
spad-lib="/usr/local/lib/fricas/target/x86_64-linux-gnu/lib/libspad.so"
foreign routines found
openServer result 0
FriCAS Computer Algebra System
Version: FriCAS 1.3.4
Timestamp: Sat Jul 14 18:19:47 EDT 2018
Ubuntu
Release 16.04.4 LTS (Xenial Xerus) 64-bit
Kernel Linux 4.4.0-130-generic x86_64
MATE 1.12.1
Oh yes, I am more that willing to change Lisp's or admit to errors.
Although being constructive is appreciated.
Ray

[Raymond Rogers]

Attached are files that illustrate the items below.  If no one objects
or corrects me I will file a bug report.
It's possible my book is incorrect.
)read test.input
Be aware that this erases the old test.log so if you want to save
previous results, rename them.
----
This item just adds to confusion
The book (?) says that
)trace )off
Kills all tracing
The attached files show that this is not presently happening
)trace name )off
stops tracing name but not other traces
In other words the tracing has to be done function by function; presently.
---
In addition
One can't trace a function until it's compiled; i.e. called once.
Which might be too late :)

Ray

Post by oldk1331
On Sun, Jul 15, 2018 at 10:58 AM, Raymond Rogers

Hi, Since there is a certain flakiness that occurs in my system. I decided
to check the installation. The compile configuration picked up sbcl 1.4.5;
The function BOOT::/UNTRACE-2 is undefined.
I recompiled and have the make file output, and, in fact, it has UNTRACE-2
as undefined. Does this indicate I have a bad lisp or some such?

It seems this function is indeed undefined, so not a lisp problem.
Can you describe how to trigger this error?

I have attached some test files that are also mysterious to me. But minor
changes change the errors and such so a large part is probably my ignorance;
sort of like quantum mechanics, the bugs are dead and alive.

In your attached file, there is an error in
facc(n) == if n < 3 then n else n * fac (n - 1)
it should be "facc" instead of "fac", is that intentional?

Misc
viewman not present, disabling graphics
/usr/local/lib/fricas/target/x86_64-linux-gnu/bin/hypertex: not found
Checking for foreign routines
AXIOM="/usr/local/lib/fricas/target/x86_64-linux-gnu"
spad-lib="/usr/local/lib/fricas/target/x86_64-linux-gnu/lib/libspad.so"
foreign routines found
openServer result 0
FriCAS Computer Algebra System
Version: FriCAS 1.3.4
Timestamp: Sat Jul 14 18:19:47 EDT 2018
Ubuntu
Release 16.04.4 LTS (Xenial Xerus) 64-bit
Kernel Linux 4.4.0-130-generic x86_64
MATE 1.12.1
Oh yes, I am more that willing to change Lisp's or admit to errors.
Although being constructive is appreciated.
Ray

[Raymond Rogers]

Found the call to UNTRACE-2
Overall examining make.lst is pretty depressing.  If nobody is going to
fix (or explain) this then I will try  another lisp.
Of course,

-----
Fricas 1.3.4

sbcl 1.4.5

--------
src/interp/slam.clisp:                     ((EQUAL |map| |x|) (IDENTITY
(/UNTRACE-2 |sub| NIL)))))))
Context:

clearCache x ==
  get(x,'localModemap,$e) or get(x,'mapBody,$e) =>
    for [map,:sub] in $mapSubNameAlist repeat
      map=x => _/UNTRACE_-2(sub,NIL)
    $e:= putHist(x,'localModemap,nil,$e)
    $e:= putHist(x,'mapBody,nil,$e)
    $e:= putHist(x,'localVars,nil,$e)
    sayKeyedMsg("S2IX0007",[x])

It would seem that whatever decides to compile functions missed this call.
Other uncompiled functions from make.lst

; compilation unit finished
;   Undefined functions:
;     /UNTRACE-2 COPY EQSUBSTLIST GENVAR HKEYS INTERNL LASSOC
LIST2REFVEC MKQ |NRTisRecurrenceRelation| SETDIFFERENCE STRINGIMAGE
|bright| |compQuietly| |eval| |get| |keyedSystemError|
|makeInternalMapMinivectorName| |mkAuxiliaryName| |mkCacheName| |pp|
|put| |putHist| |sayBrightlyI| |sayKeyedMsg| |startTimingProcess|
|stopTimingProcess| |substitute| |throwKeyedMsg| |timedOptimization|
;   Undefined variables:
;     |$InteractiveMode| |$TriangleVariableList| |$cacheAlist|
|$cacheCount| |$compileRecurrence| |$compiledOpNameList|
|$compilingInputFile| |$e| |$failed| |$functorDependencyAlist|
|$mapSubNameAlist| |$minivector| |$minivectorCode| |$minivectorNames|
|$reportCompilation|
;   caught 19 WARNING conditions
;   caught 46 STYLE-WARNING conditions
;   printed 3 notes




Ray

Post by oldk1331
On Sun, Jul 15, 2018 at 10:58 AM, Raymond Rogers

Hi, Since there is a certain flakiness that occurs in my system. I decided
to check the installation. The compile configuration picked up sbcl 1.4.5;
The function BOOT::/UNTRACE-2 is undefined.
I recompiled and have the make file output, and, in fact, it has UNTRACE-2
as undefined. Does this indicate I have a bad lisp or some such?

It seems this function is indeed undefined, so not a lisp problem.
Can you describe how to trigger this error?

[Bill Page]

It seems that UNTRACE-2 was replaced by untrace2 in this commit

https://github.com/fricas/fricas/commit/546108f1b65acadc38fc7d243b1b29770b6a770a

https://github.com/fricas/fricas/blame/546108f1b65acadc38fc7d243b1b29770b6a770a/src/interp/trace.boot#L387

but apparently this reference to UNTRACE-2 was missed.

Post by Raymond Rogers
Found the call to UNTRACE-2
Overall examining make.lst is pretty depressing. If nobody is going to
fix (or explain) this then I will try another lisp.
Of course,
-----
Fricas 1.3.4
sbcl 1.4.5
--------
src/interp/slam.clisp: ((EQUAL |map| |x|) (IDENTITY
(/UNTRACE-2 |sub| NIL)))))))
clearCache x ==
get(x,'localModemap,$e) or get(x,'mapBody,$e) =>
for [map,:sub] in $mapSubNameAlist repeat
map=x => _/UNTRACE_-2(sub,NIL)
$e:= putHist(x,'localModemap,nil,$e)
$e:= putHist(x,'mapBody,nil,$e)
$e:= putHist(x,'localVars,nil,$e)
sayKeyedMsg("S2IX0007",[x])
It would seem that whatever decides to compile functions missed this call.
Other uncompiled functions from make.lst
; compilation unit finished
; /UNTRACE-2 COPY EQSUBSTLIST GENVAR HKEYS INTERNL LASSOC LIST2REFVEC
MKQ |NRTisRecurrenceRelation| SETDIFFERENCE STRINGIMAGE |bright|
|compQuietly| |eval| |get| |keyedSystemError| |makeInternalMapMinivectorName|
|mkAuxiliaryName| |mkCacheName| |pp| |put| |putHist| |sayBrightlyI|
|sayKeyedMsg| |startTimingProcess| |stopTimingProcess| |substitute|
|throwKeyedMsg| |timedOptimization|
; |$InteractiveMode| |$TriangleVariableList| |$cacheAlist|
|$cacheCount| |$compileRecurrence| |$compiledOpNameList|
|$compilingInputFile| |$e| |$failed| |$functorDependencyAlist|
|$mapSubNameAlist| |$minivector| |$minivectorCode| |$minivectorNames|
|$reportCompilation|
; caught 19 WARNING conditions
; caught 46 STYLE-WARNING conditions
; printed 3 notes
Ray

Post by oldk1331
On Sun, Jul 15, 2018 at 10:58 AM, Raymond Rogers

Hi, Since there is a certain flakiness that occurs in my system. I decided
to check the installation. The compile configuration picked up sbcl 1.4.5;

Post by Raymond Rogers
The function BOOT::/UNTRACE-2 is undefined.

I recompiled and have the make file output, and, in fact, it has UNTRACE-2
as undefined. Does this indicate I have a bad lisp or some such?

It seems this function is indeed undefined, so not a lisp problem.
Can you describe how to trigger this error?

--
You received this message because you are subscribed to the Google Groups
"FriCAS - computer algebra system" group.
To unsubscribe from this group and stop receiving emails from it, send an
Visit this group at https://groups.google.com/group/fricas-devel.
For more options, visit https://groups.google.com/d/optout.

[Bill Page]

Post by Bill Page
It seems that UNTRACE-2 was replaced by untrace2 in this commit
https://github.com/fricas/fricas/commit/546108f1b65acadc38fc7d243b1b29770b6a770a
https://github.com/fricas/fricas/blame/546108f1b65acadc38fc7d243b1b29770b6a770a/src/interp/trace.boot#L387
but apparently this reference to UNTRACE-2 was missed.

Post by Raymond Rogers
Found the call to UNTRACE-2
Overall examining make.lst is pretty depressing. If nobody is going to fix (or explain) this then I will try another lisp.

I don't think there is anything wrong with your lisp or with your
compiled version of FriCAS.

Here is a patch for the UNTRACE-2 problem:

--

diff --git a/src/interp/slam.boot b/src/interp/slam.boot
index 3787ae0d..26208d5b 100644
--- a/src/interp/slam.boot
+++ b/src/interp/slam.boot
@@ -288,7 +288,7 @@ mkCacheVec(op,nam,kind,resetCode,countCode) ==
clearCache x ==
get(x,'localModemap,$e) or get(x,'mapBody,$e) =>
for [map,:sub] in $mapSubNameAlist repeat
- map=x => _/UNTRACE_-2(sub,NIL)
+ map=x => untrace2(sub,[])
$e:= putHist(x,'localModemap,nil,$e)
$e:= putHist(x,'mapBody,nil,$e)
$e:= putHist(x,'localVars,nil,$e)

--

You did not say specifically what you did to cause the error message
(maybe something like ')clear completely' or something like that?) so
I could not verify the fix. But at least it no longer issues that
message.

Post by Bill Page

Post by Raymond Rogers
..
It would seem that whatever decides to compile functions missed this call.

It was just an oversight.

Post by Bill Page

Post by Raymond Rogers
Other uncompiled functions from make.lst
; compilation unit finished
; /UNTRACE-2 COPY EQSUBSTLIST GENVAR HKEYS INTERNL LASSOC LIST2REFVEC MKQ |NRTisRecurrenceRelation| SETDIFFERENCE STRINGIMAGE |bright| |compQuietly| |eval| |get| |keyedSystemError| |makeInternalMapMinivectorName| |mkAuxiliaryName| |mkCacheName| |pp| |put| |putHist| |sayBrightlyI| |sayKeyedMsg| |startTimingProcess| |stopTimingProcess| |substitute| |throwKeyedMsg| |timedOptimization|
; |$InteractiveMode| |$TriangleVariableList| |$cacheAlist| |$cacheCount| |$compileRecurrence| |$compiledOpNameList| |$compilingInputFile| |$e| |$failed| |$functorDependencyAlist| |$mapSubNameAlist| |$minivector| |$minivectorCode| |$minivectorNames| |$reportCompilation|
; caught 19 WARNING conditions
; caught 46 STYLE-WARNING conditions
; printed 3 notes

As far as I can see only the first one was really a missing function.
The others are just defined in other source files.

Post by Bill Page
...

Post by Raymond Rogers

Post by oldk1331
It seems this function is indeed undefined, so not a lisp problem.
Can you describe how to trigger this error?

[Raymond Rogers]

Post by Bill Page
....
I don't think there is anything wrong with your lisp or with your
compiled version of FriCAS.

Your right; although I did try that and clisp showed the same.  gcl
failed to compile but I just let that go.

Post by Bill Page
--
diff --git a/src/interp/slam.boot b/src/interp/slam.boot
index 3787ae0d..26208d5b 100644
--- a/src/interp/slam.boot
+++ b/src/interp/slam.boot
@@ -288,7 +288,7 @@ mkCacheVec(op,nam,kind,resetCode,countCode) ==
clearCache x ==
get(x,'localModemap,$e) or get(x,'mapBody,$e) =>
for [map,:sub] in $mapSubNameAlist repeat
- map=x => _/UNTRACE_-2(sub,NIL)
+ map=x => untrace2(sub,[])
$e:= putHist(x,'localModemap,nil,$e)
$e:= putHist(x,'mapBody,nil,$e)
$e:= putHist(x,'localVars,nil,$e)
--

Thanks!

Post by Bill Page
You did not say specifically what you did to cause the error message
(maybe something like ')clear completely' or something like that?) so
I could not verify the fix. But at least it no longer issues that
message.

I'll let you know; the occurance was really hit or miss depending on
what I had in the tracing and subroutine process.  You can presume I
entered most anything including )trace )reset which is apparently
wrong.  I really thought I read that in some documentation, but I now do
)trace name )off .   Of course, then I had other problems which I
mentioned in another email. Probably the same program area as above.

....
Ray

[Raymond Rogers]

I found the particular trace sequence that aborts the calling script. 
Is this expected?
)r test-1.input
Notice that all goes well until line 15; where I try to reset the
"stats".  Which happens.
But the following code (i.e. the reading of test-1.input is aborted.
Not even hidden; after the stoppage ask:
aa
)trace
the results are in the test.log file.  So the trace is still running as
is the )spool; but test.input has stopped being read.

Ray

[Waldek Hebisch]

Post by Bill Page

Post by Bill Page
It seems that UNTRACE-2 was replaced by untrace2 in this commit
https://github.com/fricas/fricas/commit/546108f1b65acadc38fc7d243b1b29770b6a770a
https://github.com/fricas/fricas/blame/546108f1b65acadc38fc7d243b1b29770b6a770a/src/interp/trace.boot#L387
but apparently this reference to UNTRACE-2 was missed.

Post by Raymond Rogers
Found the call to UNTRACE-2
Overall examining make.lst is pretty depressing. If nobody is going to fix (or explain) this then I will try another lisp.

I don't think there is anything wrong with your lisp or with your
compiled version of FriCAS.
--
diff --git a/src/interp/slam.boot b/src/interp/slam.boot
index 3787ae0d..26208d5b 100644
--- a/src/interp/slam.boot
+++ b/src/interp/slam.boot
@@ -288,7 +288,7 @@ mkCacheVec(op,nam,kind,resetCode,countCode) ==
clearCache x ==
get(x,'localModemap,$e) or get(x,'mapBody,$e) =>
for [map,:sub] in $mapSubNameAlist repeat
- map=x => _/UNTRACE_-2(sub,NIL)
+ map=x => untrace2(sub,[])
$e:= putHist(x,'localModemap,nil,$e)
$e:= putHist(x,'mapBody,nil,$e)
$e:= putHist(x,'localVars,nil,$e)
--

Good catch, please commit.

[Raymond Rogers]

Post by Waldek Hebisch

Post by Bill Page

Post by Bill Page
It seems that UNTRACE-2 was replaced by untrace2 in this commit
https://github.com/fricas/fricas/commit/546108f1b65acadc38fc7d243b1b29770b6a770a
https://github.com/fricas/fricas/blame/546108f1b65acadc38fc7d243b1b29770b6a770a/src/interp/trace.boot#L387
but apparently this reference to UNTRACE-2 was missed.

Post by Raymond Rogers
Found the call to UNTRACE-2
Overall examining make.lst is pretty depressing. If nobody is going to fix (or explain) this then I will try another lisp.

I don't think there is anything wrong with your lisp or with your
compiled version of FriCAS.
--
diff --git a/src/interp/slam.boot b/src/interp/slam.boot
index 3787ae0d..26208d5b 100644
--- a/src/interp/slam.boot
+++ b/src/interp/slam.boot
@@ -288,7 +288,7 @@ mkCacheVec(op,nam,kind,resetCode,countCode) ==
clearCache x ==
get(x,'localModemap,$e) or get(x,'mapBody,$e) =>
for [map,:sub] in $mapSubNameAlist repeat
- map=x => _/UNTRACE_-2(sub,NIL)
+ map=x => untrace2(sub,[])
$e:= putHist(x,'localModemap,nil,$e)
$e:= putHist(x,'mapBody,nil,$e)
$e:= putHist(x,'localVars,nil,$e)
--

Good catch, please commit.

Patch worked for me.  No side effects noted running a somewhat
complicated test case that had, occasionally, thrown this error.  Of
course, if the fix merely fixed a warning/error perhaps it should still
be there in a more suitable manner.  Since that didn't happen I don't
know if I fixed the invocation or the invocation was just the result of
internal checking. (:
Ray

[Waldek Hebisch]

Post by Raymond Rogers
Found the call to UNTRACE-2
Overall examining make.lst is pretty depressing. If nobody is going to
fix (or explain) this then I will try another lisp.

Here you see "impedance mismatch". In Lisp undefined functions
and (global) variables are legal -- the assumption is that
programmer will define them as needed. Since they frequently
are errors sbcl (and probably other Lisps) warn about them.
However, Lisp do not have declaration as other languages
(Lisp "declarations" do different thing than one expects
from declarations) and I am not aware of any way to tell
Lisp compiler that function is expected to be defined
without defining it. And similarly for variables.
Consequently, warnings from Lisp do not distinguish
between functions that are not defined at all
and functions that are not defined at point of use
but are defined later. In practice almost all
warnings about undefined functions are spurious.
But _some_ warnings from sbcl proved to be quite
useful so I do not want to supress _all_ warnings.

More generaly, we get a lot of messages during build,
many of them spurious. I tried to trim them, but
in general it is hard to remove noise without
removing signal. As long as there is a lot of
noise one have to search (say using grep) for
useful info, so there is little gain from removing
only part of noise. And even spurious error
messages sometimes give useful hints (just do
not take them at face value).

[Raymond Rogers]

Post by Waldek Hebisch

Post by Raymond Rogers
Found the call to UNTRACE-2
Overall examining make.lst is pretty depressing. If nobody is going to
fix (or explain) this then I will try another lisp.

Here you see "impedance mismatch". In Lisp undefined functions
and (global) variables are legal -- the assumption is that
programmer will define them as needed. Since they frequently
are errors sbcl (and probably other Lisps) warn about them.
However, Lisp do not have declaration as other languages
(Lisp "declarations" do different thing than one expects
from declarations) and I am not aware of any way to tell
Lisp compiler that function is expected to be defined
without defining it. And similarly for variables.
Consequently, warnings from Lisp do not distinguish
between functions that are not defined at all
and functions that are not defined at point of use
but are defined later. In practice almost all
warnings about undefined functions are spurious.
But _some_ warnings from sbcl proved to be quite
useful so I do not want to supress _all_ warnings.
More generaly, we get a lot of messages during build,
many of them spurious. I tried to trim them, but
in general it is hard to remove noise without
removing signal. As long as there is a lot of
noise one have to search (say using grep) for
useful info, so there is little gain from removing
only part of noise. And even spurious error
messages sometimes give useful hints (just do
not take them at face value).

Thanks for the patch and clarification!  My only, personal apparently,
concern is that the end uses doesn't seem to be totally insulated from
the Lisp underpinnings.

Ray