Outdated msieve build in Arch AUR
 

So, I decided to take on a number to post-process for NFS@Home (namely, 8p3_870M), and I wanted to try the number out on an Arch Linux install. I saw in the AUR that there is an msieve package which can be installed as follows:

[code]pikaur -S msieve[/code]

When I do so I do get version 1.53 of msieve (with unknown svn revision). Attempting to invoke it to run the first step of post-processing, however, doesn't work:

[CODE]dylan@archlinux ~/msieve> msieve -v -nc1 "target_density=130"


Msieve v. 1.53 (SVN unknown)
Sat Sep 26 11:15:25 2020
random seeds: 2144e63a 719f4877
factoring 733642094337118721890779246550290147722658386639218063587204206764470327417566648411845811752198916045259041827753210479755559362102142022242029295265946960758537457508756105435542625633559321 (192 digits)
searching for 15-digit factors
commencing number field sieve (192-digit input)
R0: -154742504910741164739755411
R1: 126214320739011526656
A0: 16
A1: 192
A2: 0
A3: -648
A4: 144
A5: 648
A6: -270
A7: -162
A8: 81
skew 0.82, size 2.770e-06, alpha 1.432, combined = 9.417e-14 rroots = 8

commencing relation filtering
setting target matrix density to 130.0
estimated available RAM is 32067.9 MB
commencing duplicate removal, pass 1
fish: “msieve -v -nc1 "target_density=…” terminated by signal SIGSEGV (Address boundary error)
[/CODE]

At first, I thought it was a problem invoking it from the fish shell, so I tried the same thing in bash, to no avail:
[code]dylan@archlinux ~/msieve [SIGSEGV]> bash
[dylan@archlinux msieve]$ msieve -v -nc1 "target_density=130"


Msieve v. 1.53 (SVN unknown)
Sat Sep 26 11:15:45 2020
random seeds: 17e07277 26219fe0
factoring 733642094337118721890779246550290147722658386639218063587204206764470327417566648411845811752198916045259041827753210479755559362102142022242029295265946960758537457508756105435542625633559321 (192 digits)
searching for 15-digit factors
commencing number field sieve (192-digit input)
R0: -154742504910741164739755411
R1: 126214320739011526656
A0: 16
A1: 192
A2: 0
A3: -648
A4: 144
A5: 648
A6: -270
A7: -162
A8: 81
skew 0.82, size 2.770e-06, alpha 1.432, combined = 9.417e-14 rroots = 8

commencing relation filtering
setting target matrix density to 130.0
estimated available RAM is 32067.9 MB
commencing duplicate removal, pass 1
Segmentation fault (core dumped)[/code]

I tried building it again, to no avail. Eventually I solved it by compiling the latest version of msieve from subversion, which works perfectly.
Investigating further sees that the package was last updated in 2017, which is before gmp 6.2 came out, which is installed on my machine. And we know that mixing gmp versions can cause issues (see, for example, [url=https://mersenneforum.org/showthread.php?t=23087]EdH's thread on installing yafu on Ubuntu starting around post 48[/url]). So I think someone should develop a new PKGBUILD for msieve, or contact the maintainer of the package to update to the latest subversion revision of the package.

I had a similar problem on ubuntu-20.04 (again, this is the update which introduced gmp-6.2); I had to remove the lines

[code]
#if GMP_LIMB_BITS == 64
dest->_mp_d[0] = src;
dest->_mp_size = (src ? 1 : 0);
#else
[/code]

from include/gmp_xface.h to get something which worked, it sounds as if the internal representation of gmp objects may have changed

[QUOTE=Dylan14;557957]So I think someone should develop a new PKGBUILD for msieve, or contact the maintainer of the package to update to the latest subversion revision of the package.[/QUOTE]


I have proposed a diff to the [C]PKGBUILD[/C] on its [URL="https://aur.archlinux.org/packages/msieve/"]AUR page[/URL].