Re: skalibs cross-compiling fails for version >= 1.4.0

From: Laurent Bercot <ska-skaware_at_skarnet.org>
Date: Wed, 15 Jan 2014 23:32:09 +0000

  Hi Christian,
  I actually replied to this on December 22nd and was waiting
for your input...
  It looks that my e-mail never made it to the list and I didn't
notice (my client/ISP setup is definitely flakey here >.>)
  Sorry about that. Here is the e-mail I sent:

-----

> -------------%<---------------------------------------------------------
> ./compile trysizeofushort.c && ./load trysizeofushort
> echo '#define SKALIBS_SIZEOFUSHORT' `./trysizeofushort` > sizeofushort.h
> /bin/sh: ./trysizeofushort: No such file or directory
> echo '#define SKALIBS_USHORT_BITS' `expr 8 \* \`./trysizeofushort\`` >> sizeofushort.h
> /bin/sh: ./trysizeofushort: No such file or directory
> expr: syntax error
> echo '/* sysdep: sizeofushort:' `./trysizeofushort`'*/' >> sizeofushort.h
> /bin/sh: ./trysizeofushort: No such file or directory
> ./compile trysizeofuint.c && ./load trysizeofuint
> echo '#define SKALIBS_SIZEOFUINT' `./trysizeofuint` > sizeofuint.h
> /bin/sh: ./trysizeofuint: No such file or directory
> echo '#define SKALIBS_UINT_BITS' `expr 8 \* \`./trysizeofuint\`` >> sizeofuint.h
> /bin/sh: ./trysizeofuint: No such file or directory
> expr: syntax error echo '/* sysdep: sizeofuint:' `./trysizeofuint` '*/'
>>> sizeofuint.h
> /bin/sh: ./trysizeofuint: No such file or directory
>>> echo '/* ISC license. */' > sysdeps.h
> echo >> sysdeps.h
> echo '#ifndef SYSDEPS_H' >> sysdeps.h
> echo '#define SYSDEPS_H' >> sysdeps.h
> echo >> sysdeps.h
> -------------%<---------------------------------------------------------

  That is very weird. Can you go to compile/sysdeps and run by hand
./compile trysizeofushort.c
./load trysizeofushort
  to see what happens, and post the error message if any ?

  Nevertheless, it's a bug: if there's an error at those stages, the compilation
should fail instead of silently propagating the error. I'll fix the Makefile
for the next release, thanks.


> Just for the records regarding the massive change from 1.3.0 -> 1.4.0:
> ----------------------------------------------------------------------
> 357 files changed, 2302 insertions(+), 1674 deletions(-)
> ----------------------------------------------------------------------
>
> This really shows the usefulness of a public repository in my opinion
> where people could see the changes in a more atomic way and I do not
> really get why you refuse to provide something like that.

  I don't see the logical connection here. Yes, the change from 1.3.0 to
1.4.0 was massive. The file package/CHANGES sums up what I've been working
on; for people who want to dive into it, grabbing both tarballs and
performing a more or less fancy diff, as you have done, works. If there is
a problem with the latest version, the right thing is to send a bug-report
so I can fix things upstream - which is what you did, and I thank you for
that. I try to be generally reactive as far as bugfixes as concerned.

  I realize a public repository would certainly be more convenient for you
(yeah, all 5 of you people in the world interested in skalibs development).
However, the skarnet.org server, where I develop, is also a permanent
experiment and a showcase. Any software that runs on it has to:
  - be as close to bug-free as software gets
  - be lightweight (that also means C only, maybe C++ if it's otherwise perfect,
and as few dependencies as possible)
  - compile with musl and link statically without trouble
  - run its web UI, if any, with pure CGI (suitable to be served by busybox httpd).
  I'm very anal about this - it's the very point of my project in the first place,
and when I can't find software that fits my constraints, I write it. (And there's
no way I'm going to write a version control system in the next 40 years.)

  So far, git with cgit (not gitweb) almost looks like a suitable candidate. But
since I'd have to compile and set it up manually - as I do with every piece of
software running on skarnet.org - and it isn't absolutely needed, for now I have
not been prioritizing it. I can probably be talked into trying it out, if you
insist long enough.

-----

-- 
  Laurent
Received on Wed Jan 15 2014 - 23:32:09 UTC

This archive was generated by hypermail 2.3.0 : Sun May 09 2021 - 19:38:49 UTC