[PATCH] fix test failures for Clozure CL

[Qian Yun]

There was test failures for Clozure CL (CCL) before:

agcd.output: 1
integ.output: 1
lodof.output: 1
mantepse.output: 2

And new failures after recent commit:
(see https://github.com/fricas/fricas/issues/135)

fftst.output: 17

The error message is "Error reporting error" which comes from
CCL instead of FriCAS.

After some debugging, I find that following patch fixes these
test failures, which strongly suggest the root cause is inside
CCL compiler:

I just rewrite the nested assignment into two assignments.

- Qian

diff --git a/src/algebra/ffact.spad b/src/algebra/ffact.spad
index c857f48c..dc459304 100644
--- a/src/algebra/ffact.spad
+++ b/src/algebra/ffact.spad
@@ -321,7 +321,8 @@ ModularFactorizationTools1 : Exports ==
Implementation where
tmp2 := new(qcoerce(nr0)@NNI, 0)$U32Vector
for j in 0..(nc - 1) repeat
for i in 0..(nr0 - 1) repeat
- tmp2(i) := tmp1(i) := m(i, j)
+ tmp1(i) := m(i, j)
+ tmp2(i) := tmp1(i)
if j < nr1 then
rtmp := red_pol!(tmp1, rdata1)
for i in 0..(nr1 - 1) repeat

[Waldek Hebisch]

Thanks for tracking this. It would be good to report trouble to
Clozure CL folks. But should also commit the fix.

BTW: After

)set break break

Clozure CL seem to just exit on this error, that clearly is
not desired behaviour of Clozure CL (and "Error reporting error"
indicates the same).

--
Waldek Hebisch

[Qian Yun]

On 9/8/23 02:48, Waldek Hebisch wrote:
>
> Thanks for tracking this. It would be good to report trouble to
> Clozure CL folks. But should also commit the fix.

Let's see if I can get a minimal reproducible example for CCL.

> BTW: After
>
> )set break break
>
> Clozure CL seem to just exit on this error, that clearly is
> not desired behaviour of Clozure CL (and "Error reporting error"
> indicates the same).
>

Yes, I have to remove that line from CCL, then compile CCL to be
able to get backtrace for FriCAS.

- Qian