ASSERT_ALWAYS failing

[mgeisert]

Hi folks,
I'm running ecm-7.0.4 with gmp-6.2.1 on Windows 10 64-bit via Cygwin. Self-compiled ecm and gmp, details on request. I've simplified the reproducer as much as I can to get the following recipe:
ecmfactor -t 2 C1133 1e2
which reports an assertion failure on schoen_strass.c:91. C1133 is the famous unresolved portion of the 12th Fermat number. An odd thing is that specifying -t 1 works fine; it's that more than one thread provokes the error.

Has this error been reported before and/or is there a known fix?
Thanks much,

..mark

[mgeisert]

(gdb) r
Starting program: /home/Mark/src/ecm-7.0.4/.libs/ecmfactor -t 2 40386086203521847842442038116961395908045388225743593887534051882867088867451422279926227658353369309602134937818767935489955578234805439581534625494985324713549730074875381338741302421655631135507379857269344735228428553001352597596691638801743636629329355013511352942721273050339170429834278987040381747960884411851433916486144170476008852597093750739127802680309124526032940172579802008470093339990359384991503503614458710698904103258512429909701566697333753540519871100983916899540657050034590964623607736274756781417764221105569531562147057912826327014822324375878810085123801163054580870423717464500275259286644790292287618742984022979008217487409481420224445378839089353872030913057691176817044086502550278535272750787424451118761716552849620868806555149154293300951201837849814408323169458959040706805689440625767829357238882085766112685307073105174963070427573702738362159098805163558648031695168433171835632969724871377060199849541218845177450315677176119955499412825504179204908105894697957170244421770769366517833025139901383316838896066622320648993411213811241118825043233830644947321875736006536117418021434702300430337 1e2
[New Thread 5296.0x2758]
[New Thread 5296.0x4f8]
[New Thread 5296.0xe94]
[New Thread 5296.0x1c38]
Performing 2 curve(s) with B1=100
[New Thread 5296.0xfac]
[New Thread 5296.0xec4]
assertion "size <= (mp_size_t) n / (64 - 0) + 1" failed: file "schoen_strass.c", line 91, function: F_mod_1
                           [Switching to Thread 5296.0xec4]

Thread 7 "ecmfactor" hit Breakpoint 1, abort ()
    at /oss/src/winsup/cygwin/signal.cc:398
398     {
(gdb) bt
#0  abort () at /oss/src/winsup/cygwin/signal.cc:398
#1  0x00000001800434f7 in __assert_func (
    file=0x4af349875 <__func__.0+85> "schoen_strass.c", line=91,
    func=<optimized out>,
    failedexpr=0x4af349850 <__func__.0+48> "size <= (mp_size_t) n / (64 - 0) + 1") at /oss/src/winsup/cygwin/assert.cc:58
#2  0x00000004af318587 in F_mod_1 (RS=0x8000bf310, n=4096)
    at schoen_strass.c:91
#3  0x00000004af31b001 in F_toomcook4 (C=0x8000bf2c0, A=0x800089ad0,
    B=0x800089b10, len=4, n=4096, t=0x8000bf340) at schoen_strass.c:654
#4  0x00000004af31cecb in __ecm_F_mul (R=0x8000bf2c0, A=0x800089ad0,
    B=0x800089b10, len=4, parameter=1, n=4096, t=0x8000bf340)
    at schoen_strass.c:1023
#5  0x00000004af30c117 in __ecm_list_mul (a=0x8000bf2c0, b=0x800089ad0, k=4,
    c=0x800089b10, l=4, monic=1, t=0x8000bf340) at listz.c:285
#6  0x00000004af30c743 in __ecm_PolyFromRoots_Tree (G=0x800096b30,
    a=0x800096b30, k=8, T=0x8000bf2c0, dolvl=-1, n=0xffbfc490,
    Tree=0x80003a5f0, TreeFile=0x0, sh=0) at listz.c:441
#7  0x00000004af30ef92 in __ecm_stage2 (f=0x80003a960, X=0xffbfc850,
    modulus=0xffbfc930, dF=8, k=5, root_params=0xffbfc670, use_ntt=0,
    TreeFilename=0x0, stop_asap=0x0) at stage2.c:478
#8  0x00000004af305935 in ecm (f=0x80003a960, x=0x80003a980, y=0x80003a990,
    param=0x80003a9a0, sigma=0x80003a9a8, n=0x80003a948, go=0x80003a9c8,
    B1done=0x80003a9d8, B1=100, B2min_parm=0x80003a9e0, B2_parm=0x80003a9f0,
    k=5, S=0, verbose=0, repr=2, nobase2step2=0, use_ntt=1, sigma_is_A=0,
    zE=0x80003bcb0, os=0x18022eef0 <reent_data+1520>,
    es=0x18022efa8 <reent_data+1704>, chkfilename=0x0, TreeFilename=0x0,
    maxmem=0, stage1time=0, rng=0x80003aa48, stop_asap=0x0,
    batch_s=0x80003aa78, batch_last_B1_used=0x80003aa88, gw_k=0, gw_b=0,
    gw_n=0, gw_c=0) at ecm.c:1575
#9  0x00000004af322de2 in ecm_factor (f=0x80003a960, n=0x80003a948, B1=100,
    p0=0x80003a978) at factor.c:147
#10 0x00000001004010c0 in one_thread (args=0x80003a948) at ecmfactor.c:43
#11 0x0000000180155c3f in pthread::thread_init_wrapper (arg=0x80005cda0)
    at /oss/src/winsup/cygwin/thread.cc:2022
#12 0x00000001800dab4a in pthread_wrapper (arg=<optimized out>)
    at /oss/src/winsup/cygwin/miscfuncs.cc:455
#13 pthread_wrapper (arg=<optimized out>)
    at /oss/src/winsup/cygwin/miscfuncs.cc:381
#14 0x0000000000000000 in ?? ()
Backtrace stopped: previous frame inner to this frame (corrupt stack?)
(gdb)

[mgeisert]

I belatedly found where to report GMP-ECM issues: ecm-discuss at inria.fr
Thanks to Paul Zimmermann for the quick response!