Wrong result from LT for rational numbers!

[stevelinton]

Arising from a bug that Leonard Soicher had, we have a reproducible serious bug in rational number arithmetic

Observed behaviour

gap> x := -9303829819079165/4076863488;
-9303829819079165/4076863488
gap> x < x; 
false
gap> y := -9303829819079165/4076863488;
-9303829819079165/4076863488
gap> x < y;
true
gap> x = y;
true
gap> y < x;
true

Expected behaviour

x < y and y < x should return false.

Copy and paste GAP banner (to tell us about your setup)

[sal@MCC02XF17NJGH8 crypting-0.9 (master)]$ gap4
 ┌───────┐   GAP 4.10dev-2057-g8e281e4 of today
 │  GAP  │   https://www.gap-system.org
 └───────┘   Architecture: x86_64-apple-darwin18.5.0-default64-kv6
 Configuration:  gmp 6.1.2, GASMAN
 Loading the library and packages ...
 Packages:   AClib 1.3.1, Alnuth 3.1.0, AtlasRep 1.5.1, AutoDoc 2018.09.20, AutPGrp 1.10, CRISP 1.4.4, Cryst 4.1.18, CrystCat 1.1.8, 
             CTblLib 1.2.2, FactInt 1.6.2, FGA 1.4.0, GAPDoc 1.6.2, IRREDSOL 1.4, LAGUNA 3.9.0, Polenta 1.3.8, Polycyclic 2.14, 
             PrimGrp 3.3.2, RadiRoot 2.8, ResClasses 4.7.1, SmallGrp 1.3, Sophus 1.24, SpinSym 1.5, TomLib 1.2.7, TransGrp 2.0.4, 
             utils 0.59
 Try '??help' for help. See also '?copyright', '?cite' and '?authors'

I'm looking into this now.

[stevelinton]

Turns out that this is actually an integer problem

gap> q1 := -37930444087969493570027520;
-37930444087969493570027520
gap> q2 := -37930444087969493570027520;
-37930444087969493570027520
gap> q1 < q2;
true
gap> q1+1 < q2+1;
true
gap> q1+1000000 < q2+1000000;
true
gap> 2*q1 < 2*q2; 
true

[stevelinton]

Looks like the bug is in src/integer.c:1210

 /* if both arguments are negative, flip the result */
    if (IS_INTNEG(opL))
        res = !res;

This is incorrect when the two values are equal.

[stevelinton]

Seems to have been introduced in a refactoring last December.

[stevelinton]

I have a fix. Will make PR shortly.

[fingolfin]

Ouch, that's quite terrible. Glad you caught it.

[olexandr-konovalov]

This still has to be cherry-picked onto stable-4.10 and is worth making a GAP 4.10.2 release (also, about 40 package updates are waiting). I am reopening this issue to prevent it falling below the radar.

[wilfwilson]

@alex-konovalov The fix for this (PR #3478) has been backported to stable-4.10 now. Is there any reason you still want to keep this open?

[olexandr-konovalov]

The fix backported to stable-4.10 in commit 3ad3a2b