bug#12182: guile 2.0.6: "Value out of range: 0" when compiling a procedure

[Ian Price]

Feufochmar <address@hidden> writes:

>
> The backtrace I have with guild compile :
>
> $ guild compile test.scm
> Backtrace:
> In language/tree-il/optimize.scm:
>   44: 19 [optimize! #<tree-il (lambda () #)> #<module (#{ g40}#) 2097ab0> ...]
> In language/tree-il/cse.scm:
>  537: 18 [visit #<tree-il (lambda () (lambda-case #))> #<vlist ()> ...]
>  543: 17 [visit #<tree-il (lambda-case (# #))> #<vhash 22cf540 1 pairs> ...]
>  483: 16 [visit #<tree-il #> #<vhash 22cf540 1 pairs> #<vlist ()> ...]
>  537: 15 [visit #<tree-il (lambda (#) (lambda-case #))> #<vhash 22cf540 1 
> pairs> ...]
>  543: 14 [visit #<tree-il (lambda-case (# #))> #<vhash 22cf4a0 2 pairs> ...]
>  429: 13 [visit #<tree-il (let # # # ...)> #<vhash 22cf4a0 2 pairs> #<vlist 
> ()> ...]
>  370: 12 [lp (#<tree-il (lambda () #)>) () #<vlist ()>]
>  537: 11 [visit #<tree-il (lambda () (lambda-case #))> #<vhash 22cf4a0 2 
> pairs> ...]
>  543: 10 [visit #<tree-il (lambda-case (# #))> #<vhash 22cf400 3 pairs> ...]
>  437: 9 [visit #<tree-il (letrec* # # ...)> #<vhash 22cf400 3 pairs> ...]
>  533: 8 [visit #<tree-il (apply # # ...)> #<vhash 23250e0 19 pairs> ...]
>  370: 7 [lp (#<tree-il #> #<tree-il #> #<tree-il #> #<tree-il #>) () #<vlist 
> ()>]
>  533: 6 [visit #<tree-il (apply # #)> #<vhash 23250e0 19 pairs> ...]
>  370: 5 [lp (#<tree-il (if # # #)>) () #<vlist ()>]
>  403: 4 [return #<tree-il (if # # #)> #<vlist ()>]
>  333: 3 [find-dominating-lexical #<tree-il (if # # ...)> 0 ...]
>  315: 2 [unroll #<vhash 23250e0 19 pairs> 19 13]
> In ice-9/vlist.scm:
>  303: 1 [vlist-ref #<vhash 23250e0 19 pairs> 19]
> In ice-9/boot-9.scm:
>  106: 0 [#<procedure 204f280 at ice-9/boot-9.scm:97:6 (thrown-k . args)> 
> out-of-range ...]
>
> ice-9/boot-9.scm:106:20: In procedure #<procedure 204f280 at 
> ice-9/boot-9.scm:97:6 (thrown-k . args)>:
> ice-9/boot-9.scm:106:20: Value out of range: 0

I am not familiar with the cse module, but a quick perusal of the code
leads me to think that the unroll procedure in cse.scm expects the invariant

base + n <= (vlist-length db)

scheme@(guile−user) [1]> ,up
In language/tree−il/cse.scm:
   315:10  1 (unroll #<vhash 96cca60 19 pairs> 19 13)
scheme@(guile−user) [1]> ,up
In language/tree−il/cse.scm:
   333:28  2 (find−dominating−lexical #<tree−il (if (lexical steep? steep?−818) 
(le…> …)
scheme@(guile−user) [1]> ,locals
  Local variables:
  $13 = exp = #<tree−il (if (lexical steep? steep?−818) (lexical y0 y0−814) 
(lexical x…>
  $14 = effects = 0
  $15 = env = #<vhash 96cc7a0 11 pairs>
  $16 = db = #<vhash 96cca60 19 pairs>
  $17 = entry−matches? = #<procedure entry−matches? (v1 v2)>
  $18 = unroll = #<procedure unroll (db base n)>
  $19 = h = 72117960
  $20 = env−len = 11
  $21 = db−len = 19
  $22 = n = 1
  $23 = m = 16
  $24 = v = (#(#<tree−il (if (apply (primitive <) (apply (primitive values) 
(i…> …) . #)
  $25 = w = #(#<tree−il (if (apply (primitive <) (apply (primitive values) (if 
(lex…> …)
  $26 = x = 72117960
  $27 = len = 4
  $28 = w = #<tree−il (if (apply (primitive <) (apply (primitive values) (if 
(lexical …>
  $29 = w = y−step
  $30 = w = y−step−827
  $31 = w = 3
scheme@(guile−user) [1]> ,use (ice-9 vlist)
scheme@(guile−user) [1]> (vlist-ref $15 $22)
$34 = (#(#<tree−il (if (apply (primitive <) (apply (primitive values) (if 
(lexical steep? steep?−818) (lexical x0 x0−813) (lexical y0 y0−814))) (apply 
(primitive values) (if (lexical steep? steep?−818) (lexical x1 x1−815) (lexical 
y1 y1−816)))) (const 1) (const −1))> y−step y−step−827 3) . 72117960)

The only call to unroll, is (unroll db m (- db-len db-len*))
the missing db-len* is the fourth element of the vector in $34 or 3. So
the call works out at (unroll db 16 (- 19 3)) == (unroll db 16 16)
and obviously 16 + 16 > 19


Not sure if that's helped at all, but the bug intrigued me when he
brought it up on #guile yesterday

-- 
Ian Price -- shift-reset.com

"Programming is like pinball. The reward for doing it well is
the opportunity to do it again" - from "The Wizardy Compiled"