Re: seg fault building npm

[Mike Pumford <mpumford%mudcovered.org.uk@localhost>]

On 18/10/2018 23:16, Mike Pumford wrote:

> in the build directory in between the make and the make install. I'm not 
> familiar enough with the pkgsrc make targets (and paxctl) to know the 
> best way to make that happen. My guess is to do it in the preinstall 
> target. Am I right?
>
> I'm also looking at the gfind dependency as well as that looks like a 
> relatively simple makefile patch.
>
> Not sure if this will help with the npm issue but one problem at a time. :)
Okay manually doing the following before make install gets rid of the 
odd errors building node:

1. paxctl +m out/Release/node
2. substitute ! for -not in the Makefile

Not sure it actually changes anything but presumably those install steps 
do something useful. ;)

It doesn't help with the npm issue. Frustrating that the core is empty 
so I ran it under gdb and got:

Starting program: /usr/pkg/bin/node bin/npm-cli.js install 
/pkg_comp/obj/pkgsrc/lang/npm/default/marked-0.4.0 --no-global 
--no-timing --no-save
[New LWP 9]........] \ postinstall: sill install executeActions
[New LWP 11].......] \ extract:tap: verb lock using 
/root/.npm/_locks/staging-31e53ac558deaaea.lock for 
/pkg_comp/obj/pkgsrc/lang/npm/default/cli-6.4.1/node_modules/.staging
[New LWP 10]
[New LWP 8]
[New LWP 7]
[New LWP 6]
[New LWP 5]
[New LWP 4]
[New LWP 3]
[New LWP 2]

Thread 1 received signal SIGSEGV, Segmentation fault.
0x0000000000970ee8 in ecp_nistz256_points_mul ()
(gdb) bt
#0  0x0000000000970ee8 in ecp_nistz256_points_mul ()
#1  0x0000000000000000 in ?? ()
(gdb) info threads
  Id   Target Id         Frame
* 1    LWP 1             0x0000000000970ee8 in ecp_nistz256_points_mul ()
  2    LWP 9             0x000073769543e28a in poll () from 
/usr/lib/libc.so.12
  3    LWP 11            0x000073769549994a in ___lwp_park60 () from 
/usr/lib/libc.so.12
  4    LWP 10            0x000073769549994a in ___lwp_park60 () from 
/usr/lib/libc.so.12
  5    LWP 8             0x000073769543e28a in poll () from 
/usr/lib/libc.so.12
  6    LWP 7             0x000073769543ed8a in _ksem_wait () from 
/usr/lib/libc.so.12
  7    LWP 6             0x000073769549994a in ___lwp_park60 () from 
/usr/lib/libc.so.12
  8    LWP 5             0x000073769549994a in ___lwp_park60 () from 
/usr/lib/libc.so.12
  9    LWP 4             0x000073769549994a in ___lwp_park60 () from 
/usr/lib/libc.so.12
  10   LWP 3             0x000073769549994a in ___lwp_park60 () from 
/usr/lib/libc.so.12
  11   LWP 2             0x000073769543e40a in _sys___kevent50 () from 
/usr/lib/libc.so.12
(gdb)

So this is the only thread active. Digging into the nodejs tree it looks 
like its using its own in tree copy of OpenSSL 1.1.0i. Are there perhaps 
some NetBSD specific fixes that need to be applied to that distribution 
before it gets built?

I seem to recall the last openssl import (not sure what version) causing 
some upheaval in -current but I'm not sure what version that was.

Mike