On 9 April 2017 at 22:45, Bruce Dubbs <bruce.du...@gmail.com> wrote:

> Ken Moffat wrote:
>
>> Mailx is one of those packages I always build before booting a new
>> system, so that fcron can send mail to my server.  The current build
>> is using exactly the same scripts / versions as I used on another
>> machine about 4 days ago (I need a fresh, clean, system to review
>> what I'm planning re llvm-3 for rust).
>>
>> To my surprise, mailx failed:
>>
>> cc -O2 -fuse-ld=gold -g  -DMAILRC='"/etc/nail.rc"'
>> -DMAILSPOOL='"/var/mail"' -DSENDMAIL='"/usr/sbin/sendmail"'    -c
>> fio.c
>> fio.c:48:2: error: #error wordexp support is required
>>   #error wordexp support is required
>>    ^~~~~
>> fio.c:59:17: fatal error: ssl.h: No such file or directory
>>   #include <ssl.h>
>>                   ^
>>
>> Looking at the log, and comparing it to past builds, it correctly
>> noticed that I don't have NSS, but then it told me
>>
>> The following optional features are enabled:
>>   + Locale support: Printable characters depend on the environment
>>   + Multibyte character support
>>   + Character set conversion using iconv()
>>   + Automatic detection of terminal character set
>>   + Networking support (IMAP, POP3, and SMTP)
>>   + S/MIME and SSL/TLS using Network Security Services (NSS)
>> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>>   + S/MIME and SSL/TLS using OpenSSL
>>
>> That was new on this failed build, and AFAICS from fio.c it either
>> uses nss OR openssl.  Not sure about the not defined HAVE_WORDEXP
>> that caused the first error.
>>
>> Anyway, in chroot I eventually tried manually (re) running 'sh
>> ./makeconfig' and this time got the expected result.  I've now added
>> that to my script and successfully built it from the script.
>>
>> At first I had thought this might be a problem with parallel make,
>> but I don't do that for this package.
>>
>
> It's hard for me to say since openssl is always the 2nd or 3rd package
> that I build in a new system, but looking at the mailx source, it is
> looking for wordexp.h and that is installed by glibc.  Alternatively, mailx
> is doing a trivial gcc build and including that file and perhaps gcc
> failed.  That does not seem likely.


Ken, I'm tagging on to the end of this post, not because I have an answer
to your issue, but because, coincidently, I'm looking at mailx at the
moment and I can see that you are the owner of the patch.

I've found another issue which, maybe, could be added to your patch.  Mailx
is supposed to look for NSS first before falling back to OpenSSL.  I have
NSS installed and it didn't find it.  On examining the config.log I can see
that it's looking in /usr/include/ssl.h when the NSS header is actually at
/usr/include/nss/ssl.h.  In the book we create the /usr/include/nss
directory but this clearly breaks the mailx build.

Richard
-- 
http://lists.linuxfromscratch.org/listinfo/blfs-dev
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page

Reply via email to