[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: -Werror / debian/rules



Hi Steve,


On Sun, Dec 7, 2008 at 12:53 AM, Steve M. Robbins <steve@sumost.ca> wrote:
> On Fri, Dec 05, 2008 at 11:57:10AM +0100, Mathieu Malaterre wrote:
>> Steve,
>>
>>   I remove the -Werror from the debian/rules file, as it currently
>> breaks cmake configuration step. See my post:
>>
>> http://www.cmake.org/pipermail/cmake/2008-December/025839.html
>>
>> I am not sure if this breaks some kind of policy.
>
> Nope.  There's no policy (that I know of) stating that you have to be
> warning-free or use -Werror.
>
> .. but it's obviously a good idea to be warning-free ...
>
> You cmake post makes it sound like it tripped up a cmake test.  I
> wonder if there's a way to have the configuration-type checks like
> this run with blank CFLAGS and use -Werror for the main build.

Because the debian package only target debian system and thus gcc
compiler. I think you can follow the trick I provided in the cmake bug
tracker:

http://cmake.org/Bug/view.php?id=8246

However this is not very smart to deactivate builtins for the entire
project. The ideal solution would be to patch cmake to pass
-fno-builtins *only* when doing the checkfunctions exists. [I wish I
knew how autoconf people handled this case]. Does anyone knows if
AC_CHECK_FUNC handle this particular case ?


> It's great that you're taking an interest in packaging GDCM.

I am trying to learn from this particular package. Hopefully I'll be
able to provide the following: pvrg, jpegls, dicom3tools,
dcmtk-snapshot once I get more confident with the process.

> I'm a little confused by a couple of recent changelog entries,
> however.  For instance, they start with "Non-maintainer upload",
> but:
>      1. I thought you were a maintainer (?), and
>      2. as far as I can tell, gdcm has never been uploaded to Debian.

Ooops. That's a mistake, I was using dch -i as I read on some random
website and I guess it adds unwanted lines... Please revert, thanks.


> In addition, this package should be versioned as a regular (not
> "native") package, which means append a "-rel" to the version;
> i.e. use "2.0.10-1" rather than "2.0.10".

cool, thanks !

> By the way, I just realized I had done a bit of work on 2.0.10 but not
> yet committed it.  I'm doing so now, tweaking the 2.0.10-1 changelog
> entry at the same time.

Fantastic !

Thanks,
-- 
Mathieu


Reply to: