gpgme 0.3.8

Stéphane Corthésy stephane@sente.ch
Fri Jul 26 16:18:01 2002


--Apple-Mail-1--6191275
Content-Transfer-Encoding: quoted-printable
Content-Type: text/plain;
	charset=ISO-8859-1;
	format=flowed

Hi,

On Thursday, July 25, 2002, at 08:01  PM, Marcus Brinkmann wrote:

On Sun, Jul 14, 2002 at 11:40:47AM +0200, St=E9phane Corth=E9sy wrote:
I tried to compile gpgme 0.3.8 on MacOS X, but I failed to compile it
out-of-the-box :-(

I hope it works better now ;)


I can't test it now, because I can't generate the automake files on=20
MacOS X; I need to wait for gpgme 0.3.9.


Basically, there are no vasprintf() and asprintf() functions on MacOS X;
I tried using the replacement you provide (it was not part of 0.3.8
distro, but I found it in the repository), but it doesn't work: even the
test suite on vasprintf() fails:

I fixed and updated the replacement, can you please try the new=20
version?  If
that doesn't work either, it might be because of using memcpy instead
va_copy.  If that is the case, we have a problem ;) [well, nothing=20
critical,
but it would be nice to see it working now, as this is the version of
vasprintf used in libiberty, too]


It still doesn't work: memcpy() is probably the problem. va_copy()=20
doesn't exist on MacOS X.


I added a fix to the makefile that should make vasprintf part of the
distribution, and add it to the objects linked into the library, too, =
can
you please verify this?


Unable to test it until gpgme 0.3.9 goes out.


BTW, it seems to me we shouldn't add vasprintf like this, because it=20
will be
exported to applications, and that might collide with a private =
vasprintf
that is not compatible.  Well, hopefully nobody does that :)

(Note that the prototype of vasprintf() in gpgme/util.h is wrong;

Indeed.

There are also #include problems in gpgme/ath-pth.c and ath-pthread.c:
instead of #including <malloc.h> (which doesn't exist on MacOS X; we
have <sys/malloc.h>), I need to #include <stdlib.h>.

Fixed, thanks.


In fact, on MacOS X we use only ath-pthread.c, as we have pthread=20
support.


During compilation there were also some #warnings which should be
corrected:

All of them fixed, thanks.

On Thu, Jul 25, 2002 at 04:12:41PM +0200, St=E9phane Corth=E9sy wrote:
Here's a new version of the patch we use on MacOS X.


It seems you have problems with #pragma weak?  Is it not supported in=20
your
configuration?  I will have to work on this, we can probably make the=20
thread
support working on other platforms but GNU systems, but it won't be as
convenient.


That pragma is currently unknown for our compiler; it will change in one=20=

month or so, as we will use gcc 3.


Thanks for having corrected all this :-)

St=E9phane


P.-S.
Due to restrictions imposed by GPL, I will abandon helping supporting=20
GPGME on MacOS X, finding no longer reasons to do it: my main goal was=20=

to use GPGME in GPGMail, which is a (hacked) plug-in for Apple's Mail;=20=

Mail is not GPL, its sources are not available, it's delivered with=20
MacOS X; GPGMail uses a BSD-like license scheme. Till now I was=20
interfacing GPGMail to gpg through pipes, I wasn't aware that I was=20
violating the GPL (note that other software do the same, like Evolution;=20=

even rewriting their own GPGME-like clone breaks GPL, if I understand it=20=

correctly). I thought that using gpg/gpgme in another process than=20
GPGMail was not breaking GPL, and was about to use a client-server=20
pattern to workaround the process limit when using GPGME. I realize now=20=

that's it's useless, and that no un-GPL'ed application on MacOS X will=20=

ever have right to use gpg :-(
I find the GPL particularly uncomfortable, but won't discuss on it -=20
there were already many messages on that subject. In my case I was just=20=

hacking around, spending a lot of time on something that is totally=20
free, and have to stop now.

--Apple-Mail-1--6191275
Content-Transfer-Encoding: quoted-printable
Content-Type: text/enriched;
	charset=ISO-8859-1

Hi,


On Thursday, July 25, 2002, at 08:01  PM, Marcus Brinkmann wrote:


<color><param>0000,0000,DEDE</param>On Sun, Jul 14, 2002 at 11:40:47AM
+0200, St=E9phane Corth=E9sy wrote:

</color><color><param>0000,6363,1212</param>I tried to compile gpgme
0.3.8 on MacOS X, but I failed to compile it=20

out-of-the-box :-(

</color><color><param>0000,0000,DEDE</param>

I hope it works better now ;)

</color><color><param>0000,0000,0000</param>


I can't test it now, because I can't generate the automake files on
MacOS X; I need to wait for gpgme 0.3.9.



</color><color><param>0000,6363,1212</param>Basically, there are no
vasprintf() and asprintf() functions on MacOS X;=20

I tried using the replacement you provide (it was not part of 0.3.8=20

distro, but I found it in the repository), but it doesn't work: even
the=20

test suite on vasprintf() fails:

</color><color><param>0000,0000,DEDE</param>

I fixed and updated the replacement, can you please try the new
version?  If

that doesn't work either, it might be because of using memcpy instead

va_copy.  If that is the case, we have a problem ;) [well, nothing
critical,

but it would be nice to see it working now, as this is the version of

vasprintf used in libiberty, too]


</color><color><param>0000,0000,0000</param>

It still doesn't work: memcpy() is probably the problem. va_copy()
doesn't exist on MacOS X.


</color><color><param>0000,0000,DEDE</param>

I added a fix to the makefile that should make vasprintf part of the

distribution, and add it to the objects linked into the library, too,
can

you please verify this?


</color><color><param>0000,0000,0000</param>

Unable to test it until gpgme 0.3.9 goes out.


</color><color><param>0000,0000,DEDE</param>

BTW, it seems to me we shouldn't add vasprintf like this, because it
will be

exported to applications, and that might collide with a private
vasprintf

that is not compatible.  Well, hopefully nobody does that :)


</color><color><param>0000,6363,1212</param>(Note that the prototype
of vasprintf() in gpgme/util.h is wrong;

</color><color><param>0000,0000,DEDE</param>

Indeed.


</color><color><param>0000,6363,1212</param>There are also #include
problems in gpgme/ath-pth.c and ath-pthread.c:=20

instead of #including <<malloc.h> (which doesn't exist on MacOS X; we=20

have <<sys/malloc.h>), I need to #include <<stdlib.h>.

</color><color><param>0000,0000,DEDE</param>

Fixed, thanks.


</color><color><param>0000,0000,0000</param>

In fact, on MacOS X we use only ath-pthread.c, as we have pthread
support.


</color><color><param>0000,0000,DEDE</param>

</color><color><param>0000,6363,1212</param>During compilation there
were also some #warnings which should be=20

corrected:

</color><color><param>0000,0000,DEDE</param>

All of them fixed, thanks.


On Thu, Jul 25, 2002 at 04:12:41PM +0200, St=E9phane Corth=E9sy wrote:

</color><color><param>0000,6363,1212</param>Here's a new version of
the patch we use on MacOS X.

</color><color><param>0000,0000,DEDE</param>


It seems you have problems with #pragma weak?  Is it not supported in
your

configuration?  I will have to work on this, we can probably make the
thread

support working on other platforms but GNU systems, but it won't be as

convenient.

</color>


That pragma is currently unknown for our compiler; it will change in
one month or so, as we will use gcc 3.



Thanks for having corrected all this :-)


St=E9phane



P.-S.

Due to restrictions imposed by GPL, I will abandon helping supporting
GPGME on MacOS X, finding no longer reasons to do it: my main goal was
to use GPGME in GPGMail, which is a (hacked) plug-in for Apple's Mail;
Mail is not GPL, its sources are not available, it's delivered with
MacOS X; GPGMail uses a BSD-like license scheme. Till now I was
interfacing GPGMail to gpg through pipes, I wasn't aware that I was
violating the GPL (note that other software do the same, like
Evolution; even rewriting their own GPGME-like clone breaks GPL, if I
understand it correctly). I thought that using gpg/gpgme in another
process than GPGMail was not breaking GPL, and was about to use a
client-server pattern to workaround the process limit when using
GPGME. I realize now that's it's useless, and that no un-GPL'ed
application on MacOS X will ever have right to use gpg :-(

I find the GPL particularly uncomfortable, but won't discuss on it -
there were already many messages on that subject. In my case I was
just hacking around, spending a lot of time on something that is
totally free, and have to stop now.


--Apple-Mail-1--6191275--