Arch Linux User Repository

One of the makefiles tries to remove a real system file during packaging. Non-root builds where a version is already installed barf. root builds would silently remove the file. This will fix it

--- programs/Makefile.program.orig      2019-06-14 20:35:45.000000000 +0100
+++ programs/Makefile.program   2020-01-30 18:54:31.226501106 +0000
@@ -112,7 +112,7 @@

 # note: remove any old vendor file installed previously
 doinstall:: $(PROGRAM) $(CONFFILES) $(EXTRA8MAN) $(EXTRA5MAN) $(EXTRA5PROC) $(LIBFILES) $(CONFDFILES)
-       @rm -f $(FINALLIBEXECDIR)/vendor
+       @rm -f $(LIBEXECDIR)/vendor
        @mkdir -p $(PROGRAMDIR) $(MANDIR8) $(MANDIR5) $(LIBDIR) $(CONFDIR) $(CONFDDIR) $(CONFDDIR)/$(CONFDSUBDIR) $(EXAMPLECONFDIR)
        @if [ -n "$(PROGRAM)" ]; then $(INSTALL) $(INSTBINFLAGS) $(PROGRAM) $(PROGRAMDIR); fi
        @$(foreach f, $(addsuffix .8, $(PROGRAM)), \

This package is missing the following build dependencies:

• inetutils (hostname command used during build)
• xmlto (used to generate man pages)
• docbook-xsl (required to have xmlto working correctly)

The "normal" diff format isn't very good. It won't apply if everything isn't exact. The context and unified will apply. Unified are the easiest to read.

Subtracting 1 isn't enough. That may shut the compiler up but it will earn us a CVE. strncpy() does not guarantee a nul end of string. The referenced patch has a decent way to guarantee a nul but I like my way better because it also guarantees no random text can appear after the nul.

The real answer is to write a custom strncpy() that has all the behavior we want, guaranteed nul and all nul to end of buffer.

@stramaz

I have the same issue when I copy/paste it from my comment.

Here is base64 version to avoid that as I don't know how to share it in a proper way.

echo "MjMzYzIzMwo8IAkJCXN0cm5jcHkoaWZyLmlmcl9uYW1lLCBvcHRhcmcsIHNpemVvZihpZnIuaWZyX25hbWUpKTsKLS0tCj4gCQkJc3RybmNweShpZnIuaWZyX25hbWUsIG9wdGFyZywgc2l6ZW9mKGlmci5pZnJfbmFtZSktMSk7CjIzNmMyMzYKPCAJCQlzdHJuY3B5KHNoYy5jZl9uYW1lLCBvcHRhcmcsIHNpemVvZihzaGMuY2ZfbmFtZSkpOwotLS0KPiAJCQlzdHJuY3B5KHNoYy5jZl9uYW1lLCBvcHRhcmcsIHNpemVvZihzaGMuY2ZfbmFtZSktMSk7Cg==" | base64 -d > gcc9-fix.patch

the sha512sum is d4b5c8418cb623fc720d9e401cbbaf1668c172af8b5c658f95efadcce165840fd35c3f4dbb62925de73c1c3e212093e2cb427424b0db4e92ff96eb1c83cd84c4

Unfortunately the patch of @mickybart doesn't work for me...:

==> Starting prepare()...
patching file programs/tncfg/tncfg.c
Hunk #1 FAILED at 233.
Hunk #2 FAILED at 236.
2 out of 2 hunks FAILED -- saving rejects to file programs/tncfg/tncfg.c.rej
==> ERROR: A failure occurred in prepare().
    Aborting...
:: Unable to build openswan - makepkg exited with code: 4

That need to be fixed upstream but the patch should look something like that to fix the compilation:

# cat gcc9-fix.patch 
233c233
<                       strncpy(ifr.ifr_name, optarg, sizeof(ifr.ifr_name));
---
>                       strncpy(ifr.ifr_name, optarg, sizeof(ifr.ifr_name)-1);
236c236
<                       strncpy(shc.cf_name, optarg, sizeof(shc.cf_name));
---
>                       strncpy(shc.cf_name, optarg, sizeof(shc.cf_name)-1);

You can add the patch in PKGBUILD/prepare():

  # GCC 9 fix
  patch "programs/tncfg/tncfg.c" "$srcdir/gcc9-fix.patch"

It works for me but that need to be discussed with upstream team.

EDIT: similar code issue: https://github.com/intel/openlldp/issues/23

downgrading gcc to 8.3.0 didn't work at all for me. Same result as before.

Thank you!! Downgrading both gcc and gcc-libs to 8.3.0 and it worked!!

Thank you for sending me reports but since I am not able to solve it and I am not using Arch Linux as my daily driver currently it's only logical I orphan this package. PS: Maybe something to do with glibc or gcc version? Try downgrading to GCC 8.2.x and try to compile just to check if that's the issue.

Building failed with the following message:

In function 'strncpy',
    inlined from 'main' at /home/elgs/.cache/yay/openswan/src/openswan-2.6.51.5/programs/tncfg/tncfg.c:233:4:
/usr/include/bits/string_fortified.h:106:10: error: '__builtin_strncpy' specified bound 16 equals destination size [-Werror=stringop-truncation]
  106 |   return __builtin___strncpy_chk (__dest, __src, __len, __bos (__dest));
      |          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
In function 'strncpy',
    inlined from 'main' at /home/elgs/.cache/yay/openswan/src/openswan-2.6.51.5/programs/tncfg/tncfg.c:236:4:
/usr/include/bits/string_fortified.h:106:10: error: '__builtin_strncpy' specified bound 12 equals destination size [-Werror=stringop-truncation]
  106 |   return __builtin___strncpy_chk (__dest, __src, __len, __bos (__dest));
      |          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
cc1: all warnings being treated as errors
make[3]: *** [/home/elgs/.cache/yay/openswan/src/openswan-2.6.51.5/Makefile.common:21: tncfg.o] Error 1
make[3]: Leaving directory '/home/elgs/.cache/yay/openswan/src/openswan-2.6.51.5/OBJ.linux.x86_64/programs/tncfg'
make[2]: *** [/home/elgs/.cache/yay/openswan/src/openswan-2.6.51.5/programs/Makefile:59: programs] Error 1
make[2]: Leaving directory '/home/elgs/.cache/yay/openswan/src/openswan-2.6.51.5/OBJ.linux.x86_64/programs'
make[1]: *** [Makefile:10: programs] Error 1
make[1]: Leaving directory '/home/elgs/.cache/yay/openswan/src/openswan-2.6.51.5/OBJ.linux.x86_64'
make: *** [Makefile:186: programs] Error 2
==> ERROR: A failure occurred in build().
    Aborting...
Error making: openswan