-
Notifications
You must be signed in to change notification settings - Fork 245
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
NTL 10.5.0 gives issues in FLINT 2.5.2 #372
Comments
Is this on a 32 or 64 bit machine, by the way?
…On 31 August 2017 at 23:02, Julien Puydt ***@***.***> wrote:
Here is what the compilation (well, the tests after the compilation) gives:
zz_pX_to_fmpz_mod_poly....FAIL:
f_poly1 = 29 33961 30692 4297 212 19629 13852 24215 33940 11611 17373 5601
8984
21242 28859 15971 18910 3642 20984 20753 22791 3191 7697 23372 5779 14170
3157
33296 3651 27641 2928
f_poly2 = 29 33961 -246763895 -34500079 -1663877 -157797138 -111344267
-1946741
98 -272876656 -93313217 -139664220 -44992724 -72192102 -170768627
-232026654 -12
8390570 -152024487 -29236779 -168697264 -166829640 -183230765 -25603403
-6183528
4 -187882841 -46418908 -113891024 -25331749 -267715228 -29304692
-222213143 -234
98084
zz_pE_to_fq....FAIL:
p = 127
mod = x^9+55*x^8+44*x^7+79*x^6+65*x^5+16*x^4+17*x^3+118*x^2+124*x+43
f1 = 9*a^8+90*a^7+12*a^6+40*a^5+108*a^4+73*a^3+28*a^2+16*a+13 - 9
zzpe:[-1384 -1762 -3274 -8944 -13354 -4786 -1258 -11086 -880]
f2 = -880*a^8-11086*a^7-1258*a^6-4786*a^5-13354*a^4-8944*a^3-3274
*a^2-1762*a-138
4 - 9
zz_pEX_to_fq_poly....Aborted
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#372>, or mute the thread
<https://github.com/notifications/unsubscribe-auth/AAOzpur9TFCZ_McbYkuJXqOHTazhdqZPks5sdx9qgaJpZM4PJZ6H>
.
|
The architecture is amd64. |
Similar failures on a 32-bit x86 Debian userland (64-bit x86-64 kernel):
|
My guess is you didn't configure NTL correctly to use GMP as its large integer package. This causes the integer format to be incompatible with Flint. |
Our configuration is here: https://anonscm.debian.org/cgit/debian-science/packages/libntl.git/tree/debian/rules#n22 Does that look wrong? NTL_GMP_LIP is actually not a typo (for LIB), LIP is correct and appears many times in the sources. |
I don't know, but I don't see where you specify where GMP is installed. Is
it installed in /usr/local/ on the machine in question?
Currently when I install NTL it says that NTL_GMP_LIP is not set, even
though I am passing that option to NTL's configure. So apparently some
additional preconditions need to be satisfied.
…On 9 October 2017 at 16:25, Ximin Luo ***@***.***> wrote:
Our configuration is here: https://anonscm.debian.org/
cgit/debian-science/packages/libntl.git/tree/debian/rules#n22
Does that look wrong? NTL_GMP_LIP is actually not a typo (for LIB), LIP is
correct and appears many times in the sources.
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#372 (comment)>, or mute
the thread
<https://github.com/notifications/unsubscribe-auth/AAOzpgm9NOdPQYbgj2KWv6UYPbxMZQZOks5sqizxgaJpZM4PJZ6H>
.
|
This is Debian so everything is installed into system locations (
We've also been building NTL and Flint with these flags since yonks, FLINT 2.5.2 works for NTL 9.9.1: https://buildd.debian.org/status/fetch.php?pkg=flint&arch=amd64&ver=2.5.2-15&stamp=1485219493&raw=0 But fails for NTL 10.5.0: https://people.debian.org/~infinity0/sage/flint_amd64-2017-08-31T12:19:51Z.build (1.8 MB) |
According to the NTL installation page, it looks in /usr/local, not /usr. I
don't know if this is relevant.
Perhaps they changed the integer format in NTL 10. That's another
possibility.
…On 9 October 2017 at 16:45, Ximin Luo ***@***.***> wrote:
This is Debian so everything is installed into system locations (/usr/lib
and /usr/lib/$(multiarch)) and is found automatically. My NTL build log
<https://people.debian.org/%7Einfinity0/sage/ntl_10.5.0-1_amd64-2017-08-31T10:39:20Z.build>
confirms the flag is set:
[..]
make[1]: Leaving directory '/<<PKGBUILDDIR>>/src'
./gen_gmp_aux > ../include/NTL/gmp_aux.h
NTL_GMP_LIP flag set
NTL_ZZ_NBITS = 64
[..]
We've also been building NTL and Flint with these flags since yonks, FLINT
2.5.2 works for NTL 9.9.1:
https://buildd.debian.org/status/fetch.php?pkg=flint&
arch=amd64&ver=2.5.2-15&stamp=1485219493&raw=0
But fails for NTL 10.5.0:
https://people.debian.org/~infinity0/sage/flint_amd64-
2017-08-31T12:19:51Z.build (1.8 MB)
https://people.debian.org/~infinity0/sage/flint_i386-
2017-09-01T08:53:17Z.build (1.9 MB)
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#372 (comment)>, or mute
the thread
<https://github.com/notifications/unsubscribe-auth/AAOzpkmQIAeWYH12Bw3AvkHF6_fllvi1ks5sqjGKgaJpZM4PJZ6H>
.
|
According to |
Are you having problems reproducing this issue? |
Yes, I cannot replicate this. Here is my configuration (note I build at a
prefix not system wide, and I am using the latest Flint master, not 2.5.2,
though the only relevant change was the order of some items when doing the
NTL interface build [1]):
NTL:
./configure DEF_PREFIX=/home/wbhart/ PREFIX=/home/wbhart/ NTL_GMP_LIP=on
SHARED=on
Flint:
./configure --enable-shared --disable-static --with-gmp=/home/wbhart
--with-mpfr=/home/wbhart --with-ntl=/home/wbhart --prefix=/home/wbhart
For some reason it didn't find the symbols in the library without setting
LD_LIBRARY_PATH before doing make check.
Naturally GMP must be built with the --enable-cxx option.
Bill.
[1] https://github.com/wbhart/flint2/pull/377/files
…On 9 October 2017 at 17:01, Ximin Luo ***@***.***> wrote:
Are you having problems reproducing this issue?
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#372 (comment)>, or mute
the thread
<https://github.com/notifications/unsubscribe-auth/AAOzpmabOnSKNMs2wtJcy49T7Fb2rd_Iks5sqjVcgaJpZM4PJZ6H>
.
|
Very strange, I can reproduce the bug no problem by building our Debian NTL 10.5.0 package using our Hopefully you can confirm this for yourself using the binaries I built. There is a
The bad one was compiled inside sbuild, the good one was compiled on my host system. I've uploaded a diffoscope analysis of the diff between the two binaries but haven't looked at it in detail yet. I will also need to figure out what the differences between sbuild vs host system is. Toolchain programs being at different versions is the obvious culprit - my host system is "Debian testing" which is older than what sbuild runs ("Debian unstable") - however the failures that @SnarkBoojum and I reported originally were probably also using older toolchain programs as well. |
It looks like it might be due to the Debian maintainer from 6 years ago deciding to set However I wasn't able to reproduce the FLINT test failures by giving MAKEFLAGS=-j4 to upstream NTL. The Debian packaging files has some too-smart-for-its-own-good extra Makefile rules to avoid running the wizard, so I suspect that's what's to blame. ("Recursive Make considered harmful"). Not sure about the NTL from 6 years ago, but today we can just set So I'll fix up the Debian packaging and run some more tests to confirm there's no problem in FLINT. Sorry for the noise and thanks for your patience. |
Unable to replicate. |
Here is what the compilation (well, the tests after the compilation) gives:
zz_pX_to_fmpz_mod_poly....FAIL:
f_poly1 = 29 33961 30692 4297 212 19629 13852 24215 33940 11611 17373 5601 8984
21242 28859 15971 18910 3642 20984 20753 22791 3191 7697 23372 5779 14170 3157
33296 3651 27641 2928
f_poly2 = 29 33961 -246763895 -34500079 -1663877 -157797138 -111344267 -1946741
98 -272876656 -93313217 -139664220 -44992724 -72192102 -170768627 -232026654 -12
8390570 -152024487 -29236779 -168697264 -166829640 -183230765 -25603403 -6183528
4 -187882841 -46418908 -113891024 -25331749 -267715228 -29304692 -222213143 -234
98084
zz_pE_to_fq....FAIL:
p = 127
mod = x^9+55x^8+44x^7+79x^6+65x^5+16x^4+17x^3+118x^2+124x+43
f1 = 9a^8+90a^7+12a^6+40a^5+108a^4+73a^3+28a^2+16a+13 - 9
zzpe:[-1384 -1762 -3274 -8944 -13354 -4786 -1258 -11086 -880]
f2 = -880a^8-11086a^7-1258a^6-4786a^5-13354a^4-8944a^3-3274a^2-1762a-138
4 - 9
zz_pEX_to_fq_poly....Aborted
The text was updated successfully, but these errors were encountered: