Re: crosstest compile fix (another)
On Sunday 22 February 2004 15:17, Stefan Leichter wrote:
as pointed out from Hans Leidekker (http://www.winehq.org/hypermail/ wine-devel/2004/02/0448.html) mingw misses some GUID in its header files
Well, it was a patch to MinGW and I don't think we should work around this shortcoming of MinGW by patching Wine. On the other hand, MinGW is still not useful for us even if we're sending them patches to fix things every now and then. That, and the poor or outdated state of MinGW RPM packages, and the plans for a PE build of Wine lead me to ask myself the question 'Why don't we build our own MinGW cross compiler packages?' That way we could have a working MinGW on RPM based distros as well and we can throw in couple of patches of our own to make it useful for compiling Wine. Not to create a fork of course, but more like what Redhat does, i.e. still trying to get patches accepted upstream. I have been working on these RPMS already and plan on putting up a page somewhere shortly. Initial goal is to compile all Wine tests out of the box. -Hans
Hans Leidekker wrote:
On the other hand, MinGW is still not useful for us even if we're sending them patches to fix things every now and then. That, and the poor or outdated state of MinGW RPM packages, and the plans for a PE build of Wine lead me to ask myself the question 'Why don't we build our own MinGW cross compiler packages?'
Amen. Not everyone need to "make crosstest". We would come a long way if testers instead could install the Windows service (not yet written) that downloads the latest crosstest EXE from winehq and runs it. regards, Jakob
Yep, should have the first version of this service this week sometime. Thanks, Chris
Amen. Not everyone need to "make crosstest". We would come a long way if testers instead could install the Windows service (not yet written) that downloads the latest crosstest EXE from winehq and runs it.
regards, Jakob
Good! I was thinking about making time for writing it, nice that I don't have to. :-) Chris Morgan wrote:
Yep, should have the first version of this service this week sometime.
Thanks, Chris
Amen. Not everyone need to "make crosstest". We would come a long way if testers instead could install the Windows service (not yet written) that downloads the latest crosstest EXE from winehq and runs it.
regards, Jakob
On Monday 23 February 2004 04:16, Hans Leidekker wrote:
I have been working on these RPMS already and plan on putting up a page somewhere shortly. Initial goal is to compile all Wine tests out of the box.
As promised, a set of MinGW RPM packages for a MinGW cross compiler that can compile all Wine tests. I've thrown in three patches of mine: one that adds missing DirectX guids, an urlmon import lib and some corrections to comctl32. Find them here: http://mirzam.it.vu.nl/mingw/ -Hans
Am Dienstag, 24. Februar 2004 17:12 schrieb Hans Leidekker:
On Monday 23 February 2004 04:16, Hans Leidekker wrote:
I have been working on these RPMS already and plan on putting up a page somewhere shortly. Initial goal is to compile all Wine tests out of the box.
As promised, a set of MinGW RPM packages for a MinGW cross compiler that can compile all Wine tests. I've thrown in three patches of mine: one that adds missing DirectX guids, an urlmon import lib and some corrections to comctl32.
Find them here:
-Hans Hello Hans,
your gcc package (gcc-mingw-3.2.3-7hl.src.rpm) does not build for me on SuSE 9.0 with the mingw packages provided from Michael Stefaniuc (Redhat: mingw-3.0-1, mingw-binutils-2.13.90.20030111.1-2, mingw-gcc-3.2.3.20030504.1-1). The problem is that your package expects mingw install in /usr/local/ i386-mingw32 but the Redhat one is in /usr/mingw. The error message is attached. Bye Stefan
On Wednesday 25 February 2004 15:06, Stefan Leichter wrote:
The problem is that your package expects mingw install in /usr/local/ i386-mingw32 but the Redhat one is in /usr/mingw. The error message is attached.
There are two ways out here. One is to change the install path to /usr/mingw in the spec file of gcc-mingw. I'm not going to do that myself because I think /usr/local/i386-mingw is a better choice (gcc-mingw is not part of the core distribution and old Unix wisdom says to put such stuff in /usr/local, i386-mingw is the official gcc target). The other way out for you is to install my ming-runtime package as I describe on my page that has the stuff in the right place to build gcc-mingw as is. By the way, if you succeed in building SUSE 9 RPMS I'm interested in putting those on my page. Could you send me the packages? -Hans
Am Donnerstag, 26. Februar 2004 03:45 schrieb Hans Leidekker:
On Wednesday 25 February 2004 15:06, Stefan Leichter wrote:
The problem is that your package expects mingw install in /usr/local/ i386-mingw32 but the Redhat one is in /usr/mingw. The error message is attached.
There are two ways out here. One is to change the install path to /usr/mingw in the spec file of gcc-mingw. I'm not going to do that myself because I think /usr/local/i386-mingw is a better choice (gcc-mingw is not part of the core distribution and old Unix wisdom says to put such stuff in /usr/local, i386-mingw is the official gcc target).
The other way out for you is to install my ming-runtime package as I describe on my page that has the stuff in the right place to build gcc-mingw as is.
I am sorry to tell you that the second way does not work for me. The error message is attached (mingw.log). So i switched back to the Redhat version and did cd /usr/local ; ln -s /usr/mingw i386-mingw32 and in /usr/bin i need the links ln -s mingw-ar i386-mingw32-ar ln -s mingw-ranlib i386-mingw32-ranlib Therefore i think the binutils package must be installed before building the gcc package. Than i removed the Redhat packages and installed your runtime and my binutils and gcc package and stat'rted building mingw-3.2. This fails with a linker error (attached: user.log). I will try to build mingw-3.2 in the evening. Bye Stefan
On Thursday 26 February 2004 00:06, Stefan Leichter wrote:
I am sorry to tell you that the second way does not work for me. The error message is attached (mingw.log). So i switched back to the Redhat version and did cd /usr/local ; ln -s /usr/mingw i386-mingw32 and in /usr/bin i need the links ln -s mingw-ar i386-mingw32-ar ln -s mingw-ranlib i386-mingw32-ranlib
Strange, I don't need those links. Is it only *-ar and *-ranlib or all of them?
Therefore i think the binutils package must be installed before building the gcc package.
Just checked that and yes, you're right. I've changed the web page and spec file accordingly.
Than i removed the Redhat packages and installed your runtime and my binutils and gcc package and stat'rted building mingw-3.2. This fails with a linker error (attached: user.log).
It's the circular dependency again: you need to have MinGW installed in order to build it. I thought it was only the mingw-runtime package but it appears you need to have mingw-w32api installed as well. I've updated the web page to reflect this. This is a problem with the upstream sources by the way. Maybe we can solve this by splitting them up and reordering them a bit, don't know, but I will look into this.
I will try to build mingw-3.2 in the evening.
Keep me posted. -Hans
participants (4)
-
Chris Morgan -
Hans Leidekker -
Jakob Eriksson -
Stefan Leichter