Arch|NetBSD: Failure with undefined BIGNUMBER when using system WolfSSL

[rdmark]

When building with system WolfSSL on Arch and NetBSD, we get errors like:

[171/254] Compiling C object etc/uams/uams_randnum.so.p/uams_randnum.c.o
In file included from ../include/atalk/logger.h:5,
                 from ../etc/uams/uams_randnum.c:36:
./config.h:378:9: warning: "HAVE_PTHREAD" redefined
  378 | #define HAVE_PTHREAD 1
      |         ^~~~~~~~~~~~
In file included from ../etc/uams/uams_randnum.c:30:
/usr/include/wolfssl/options.h:132:9: note: this is the location of the previous definition
  132 | #define HAVE_PTHREAD
      |         ^~~~~~~~~~~~
[172/254] Compiling C object etc/uams/uams_dhx_passwd.so.p/uams_dhx_passwd.c.o
FAILED: etc/uams/uams_dhx_passwd.so.p/uams_dhx_passwd.c.o 
cc -Ietc/uams/uams_dhx_passwd.so.p -Ietc/uams -I../etc/uams -I. -I.. -Iinclude -I../include -Isys -I../sys -Ietc/afpd -I../etc/afpd -fdiagnostics-color=always -D_FILE_OFFSET_BITS=64 -Wall -Winvalid-pch -Wextra -Wpedantic -std=c11 -O0 -g -DHAVE_CONFIG_H '-D_U_=__attribute__((unused))' -Wno-pedantic -Wno-extra -Wno-all -Wno-deprecated-declarations -D_GNU_SOURCE -fPIC -MD -MQ etc/uams/uams_dhx_passwd.so.p/uams_dhx_passwd.c.o -MF etc/uams/uams_dhx_passwd.so.p/uams_dhx_passwd.c.o.d -o etc/uams/uams_dhx_passwd.so.p/uams_dhx_passwd.c.o -c ../etc/uams/uams_dhx_passwd.c
../etc/uams/uams_dhx_passwd.c: In function ‘pwd_login’:
../etc/uams/uams_dhx_passwd.c:81:5: error: unknown type name ‘BIGNUM’
   81 |     BIGNUM *bn, *gbn, *pbn;
      |     ^~~~~~
../etc/uams/uams_dhx_passwd.c:82:11: error: unknown type name ‘BIGNUM’
   82 |     const BIGNUM *pub_key;
      |           ^~~~~~

[rdmark]

The two cases are quite different.

On NetBSD the build system actually warns that WolfSSL doesn't have all the symbols and flag SSL as built-in. However it seems like a macro gets set incorrectly and the wrong headers get included.

On Arch, the build system things WolfSSL is good and flags SSL as WolfSSL. But in reality capabilities are missing.

[rdmark]

After https://github.com/Netatalk/netatalk/pull/1440 the NetBSD issue got resolved. Probably because we got rid of all of the local macro hacks.

We are still getting missing symbol errors on Arch.

[VorpalBlade]

Has the issue been reported in the Arch Linux bug tracker, over on their gitlab?

[rdmark]

It has not, to my best knowledge. But it’s a good idea. Shall I give it a go or are you volunteering? :)

[VorpalBlade]

You are probably more knowledgeable on what the actual specific issue is than I am. I'm just a C++ and Rust programmer with rusty (pun not intended) C knowledge, who don't really know these libraries at all.

https://archlinux.org/packages/extra/x86_64/wolfssl/ has the relevant links to the gitlab. If you can't get an account on their bug tracker (process a bit manual due to spam issues, but I think you can log in with github to get around that) I can absolutely report this, but you need to provide some details so I understand the situation fully.

[rdmark]

The Arch anti spam filter defeated me a few months ago but I can try again with your advice.

Edit: Sent an account creation request to the given email address.

[VorpalBlade]

Did you manage to get through? Or should I report it using my existing account?

[rdmark]

Reported! https://gitlab.archlinux.org/archlinux/packaging/packages/wolfssl/-/issues/2

At a related note, there's this bug too: https://gitlab.archlinux.org/archlinux/packaging/packages/wolfssl/-/issues/1

So it seems that the Arch wolfSSL not being full-featured is a known issue.