- From: Regis Boudin <regis@boudin.name>
- Date: Wed, 9 Jan 2008 10:26:18 -0000 (GMT)
- To: www-amaya@w3.org
On Wed, January 9, 2008 08:46, Matej Cepl wrote: > > On 2008-01-08, 16:33 GMT, Regis Boudin wrote: >> -------- >> + make -j3 all >> -------- >> I would avoid that. Someone at Debian tried to rebuild the whole archive >> with -j2, it appears that Amaya doesn't cope well with it. :( > > That's Fedora/Red Hat default -- we just seem to trust in the > quality of our upstream authors too much ;-). Could fix, but > I was able to build amaya with it, so I will keep it present for > now. I would say it's mainly a matter of luck. Our test with -j3 failed with "/usr/bin/ld: final link failed: Text file busy". >> See the private mail I sent you earlier for details, but you >> got it pretty >> much right. > > As a former Debian user, I can still read debian/rules pretty > well ;-). Well, the thing is, patches are not necessarilly in a patches/ directory, I only keep them here for reference, and to make it easier for other people to understand what is actually patched. >>> 2) When bin/amaya_bin is being linked (line 2919 and its >>> surroundings) linking crashes (version of system libwww is >>> 5.4.1): >>> >>> query.o: In function `AHTProfile_newAmaya': >>> /home/matej/redhat/BUILD/Amaya/WX/amaya/../../amaya/query.c:2430: >>> undefi ned reference to `HTTransportInit()' >>> /home/matej/redhat/BUILD/Amaya/WX/amaya/../../amaya/query.c:2469: >>> undefined reference to `HTTransferEncoderInit(_HTList*)' >>> /home/matej/redhat/BUILD/Amaya/WX/amaya/../../amaya/query.c:2487: >>> undefined reference to `HTMIMEInit()' >>> collect2: ld returned 1 exit status >>> > I was trying to understand query.c and I have to admit that > although I am not a programmer, I didn't like what I saw. Note, > that it doesn't ask for HTTransportInit, which is really in > libwwwapp, but for the C++ object HTTransportInit() (and later > HTTransportInit(_HTList*) -- when you saw last asterisk in > C symbol?). This is actually the linker saying the function it looks foor takes a pointer to _HTList as a parameter, nothing wrong here. > It seems to me that the author of query.c (and > related files) got somehow really confused in dealing with > C symbols inside of C++ files and the result is mess -- it seems > to me that the particular symbols are generated twice inline and > the system one in libwwwapp is never used anyway. After all this > struggle I have created this brutal patch which made Amaya to > compile: Though I agree that query.c looks messy, I'm not sure the problem is related to the C/C++ confusion. I hadn't noticed that Amaya embedded more copies of files from libwww... But they aren't compiled anyway, so I have no idea how your patch fixed the issue :/ > But I am afraid that means that the internal libwww is used after > all, right? At this point I have no idea. A build log would have the answer ;) But if you still had my patches applied and used the --enable-system-libwww, probably not. You might have compiled/used an internal copy of one file. > I was able to compile Amaya with it -- whoever > wanted for the Fedora Amaya, it is on > http://mcepl.fedorapeople.org/rpms/amaya-9.55.2-1.fc8.i386.rpm > but it doesn't work wek well -- the problems with display of the > webpages is constant. There are some issues with display and DRI on some video cards, might be that. >>> and (I guess) >>> one of the reasons why many distros (aside from the ones >>> managed by maniacs like Debian Developers -- that's meant as >>> a praise) >> >> I'll take that as a compliment, then. :) > > Yes, s/praise/compliment/ -- my English betrayed me ;-). I was not pedantic picky about vocabulary, don't worry. Just commenting on the fact that I am described as "maniac" ;) Regis
Received on Wednesday, 9 January 2008 10:26:38 UTC