- Dillo-dev - lists.mailman3.com

Re: [Dillo-dev]AIX : illegal instruction

by Andreas Schweitzer

Hi, I followed the progress on the mailing list, but was always too late ;-) ... About the WAIT_ANY : AIX 4.3.3 (and Ricardo's 4.2.X) indeed does not define WAIT_ANY. I checked man-pages on several OS's and neither says anything about this being standard, but all agree that "-1" is standard, so "-1" is probably the way to go. > I compiled dillo 0.80 Jan-30 tarball after having applied the patches by > Andreas (namely undefining func_data before ever gtk.h include) I never dug deep enough to find out if this is the right way to solve this problem. Apparently, both AIX-internals and GTK define this "func_data", I only remember that the AIX definition was for something we don't need - but I don't remember the details. > > % grep func_data * > browser.h:#undef func_data > dns.h:#undef func_data > web.h:#undef func_data *If* that ever wants to go into dillo, a cleaner way would be to enclose it in "#ifdef _AIX" - but I don't know about AIX5 > The compilation ends (although I saw some wanrings about vsnprintf > implicitly declared... > > But startup fails, GDB gives me the following: > (gdb) bt > #0 0x0 in ?? () from /usr/local/build/dillo-0.8.0-pre/src/./dillo and now I'm lost ... is this all the output ? Cheers, Andreas -- **************************** NEW ADDRESS ****************************** Hamburger Sternwarte Universitaet Hamburg Gojenbergsweg 112 Tel. ++49 40 42891 4016 D-21029 Hamburg, Germany Fax. ++49 40 42891 4198

Feb. 1, 2004

[Dillo-dev]AIX : illegal instruction

by Riccardo

Hello, I compiled dillo 0.80 Jan-30 tarball after having applied the patches by Andreas (namely undefining func_data before ever gtk.h include) % grep func_data * browser.h:#undef func_data dns.h:#undef func_data web.h:#undef func_data The compilation ends (although I saw some wanrings about vsnprintf implicitly declared... But startup fails, GDB gives me the following: (gdb) bt #0 0x0 in ?? () from /usr/local/build/dillo-0.8.0-pre/src/./dillo cheers, Riccardo

Feb. 1, 2004

Re: [Dillo-dev] building on AIX, gcc, dillo 0.7.3

by Jorge Arellano Cid

On Sat, 31 Jan 2004, Riccardo wrote: > > > > What could be the problem? > > > > A missing header! Fix is commited. > > Please test it and tell us how it works. > > > > -- > > To test, I had to use the CVS of course. I run autogen.sh and > > if gcc -DHAVE_CONFIG_H -I. -I. -I.. > -DDPIDRC_SYS='"/usr/local/etc/dpidrc"' -I/usr/local/include > -I/usr/local/include/glib-1.2 -I/usr/local/lib/glib/include -g -O2 > -D_REENTRANT -D_THREAD_SAFE -Wall -W -Wno-unused-parameter > -Waggregate-return -MT dpid.o -MD -MP -MF ".deps/dpid.Tpo" \ > -c -o dpid.o `test -f 'dpid.c' || echo './'`dpid.c; \ > then mv ".deps/dpid.Tpo" ".deps/dpid.Po"; \ > else rm -f ".deps/dpid.Tpo"; exit 1; \ > fi > dpid.c: In function `handle_sigchld': > dpid.c:562: error: `WAIT_ANY' undeclared (first use in this function) > dpid.c:562: error: (Each undeclared identifier is reported only once > dpid.c:562: error: for each function it appears in.) > make[2]: *** [dpid.o] Error 1 Hmmm.... > Is there still something missing in cvs? WHy is it not there? Maybe > release me a new tarball so I can test it on both aix and sparc, or tell > me how to build cleanly from CVS. > Or WAIT_ANY is not declared on solaris? hmm :-) Some Solaris seem not to have it. :( I decided to replace WAIT_ANY with -1 as in dillo.c; commited. Please test it on aix and sparc. Use the Jan30 tarball and just replace WAIT_ANY with -1 in: dpi/downloads.c and dpid/dpid.c Cheers Jorge.- PS: there's no need to CC me. I'm on dillo-dev.

Feb. 1, 2004

Re: [Dillo-dev] building on AIX, gcc, dillo 0.7.3

by Riccardo

> > What could be the problem? > > A missing header! Fix is commited. > Please test it and tell us how it works. > > -- To test, I had to use the CVS of course. I run autogen.sh and if gcc -DHAVE_CONFIG_H -I. -I. -I.. -DDPIDRC_SYS='"/usr/local/etc/dpidrc"' -I/usr/local/include -I/usr/local/include/glib-1.2 -I/usr/local/lib/glib/include -g -O2 -D_REENTRANT -D_THREAD_SAFE -Wall -W -Wno-unused-parameter -Waggregate-return -MT dpid.o -MD -MP -MF ".deps/dpid.Tpo" \ -c -o dpid.o `test -f 'dpid.c' || echo './'`dpid.c; \ then mv ".deps/dpid.Tpo" ".deps/dpid.Po"; \ else rm -f ".deps/dpid.Tpo"; exit 1; \ fi dpid.c: In function `handle_sigchld': dpid.c:562: error: `WAIT_ANY' undeclared (first use in this function) dpid.c:562: error: (Each undeclared identifier is reported only once dpid.c:562: error: for each function it appears in.) make[2]: *** [dpid.o] Error 1 Is there still something missing in cvs? WHy is it not there? Maybe release me a new tarball so I can test it on both aix and sparc, or tell me how to build cleanly from CVS. Or WAIT_ANY is not declared on solaris? hmm :-) -Ric

Jan. 31, 2004

Re: [Dillo-dev] building on AIX, gcc, dillo 0.7.3

by Jorge Arellano Cid

On Sat, 31 Jan 2004, Peter Sinn wrote: > > Hi, > > first let me say that I have been using Dillo on my Sun Ultra 10 for > quite some time now and like it very much! Thanks for the comment Peter! > Am 31. January 2004 12:14 Uhr schrieb Jorge Arellano Cid: > > > There's no PF_LOCAL in the CVS tarball. Please try it. > > > > (http://www.dillo.org/test/dillo-0.8.0-pre.Jan30.tar.gz) > > > > dillo 7.3 compiles without problems on my machine, therefore I have > just tried to compile the new version. But unfortunately, with > dillo-0.8.0-pre.Jan30 I get the following compile error (gcc version > 3.3.2, SPARC/Solaris 9): > > if gcc -DHAVE_CONFIG_H -I. -I. -I.. > -DDPIDRC_SYS='"/usr/local/etc/dpidrc"' -I/usr/local/include > -I/usr/local/include/glib-1.2 -I/usr/local/lib/glib/include -g -O2 > -D_REENTRANT -D_THREAD_SAFE -Wall -W -Wno-unused-parameter > -Waggregate-return -MT dpid.o -MD -MP -MF ".deps/dpid.Tpo" \ > -c -o dpid.o `test -f 'dpid.c' || echo './'`dpid.c; \ > then mv -f ".deps/dpid.Tpo" ".deps/dpid.Po"; \ > else rm -f ".deps/dpid.Tpo"; exit 1; \ > fi > dpid.c: In function `handle_sigchld': > dpid.c:561: error: `WAIT_ANY' undeclared (first use in this function) > dpid.c:561: error: (Each undeclared identifier is reported only once > dpid.c:561: error: for each function it appears in.) > make[2]: *** [dpid.o] Error 1 > > > What could be the problem? A missing header! Fix is commited. Please test it and tell us how it works. -- Ah, I also fixed the handling of the <P> element for the bug-meter. It fixes the problem of wrong messages like this one: "HTML warning: line 200, unexpected closing tag: </p>. -- expected </ol>" (now this part should be release-ready) -- Thanks for the quick feedback Jorge.- PS: There's no need to CC me.

Jan. 31, 2004

Re: [Dillo-dev] Automake and dynamic linking

by robert of northworthige

In message <Pine.LNX.4.58.0401301144290.732(a)infinity.cl>, Jorge Arellano Cid <jcid(a)dillo.org> writes > >Jorgen, >Andreas, >Riccardo, >Matti, >Bob >(portability issues here): > snip >PS2: >--- >> [Bob wrote] >> - my experience of recent cvs, compiled on the playstion2, (blackrhino >> 2.2.21 kernel), apparently shows a disconcerting feature. After a few >> minutes of scrolling up and down standard web-sites the whole system >> will hang (eg it's unpingable, and doesn't respond to any keyboard >> input), and needs a hard reset :-( >> >> Anyone else currently testing on the ps2 please (and hopefully had >> better experiences, or a fix)?? > > I don't have a PS2, but when testing on libc5, there was a >similar problem (crash after scrolling, or crash instants after >starting) when _some_ of the libraries used by the X server where >compiled reentrant and some others not. In fact I was upgrading >slk3.6's with the ones at: http://www.linuxsavvy.com/resources/ >and some links pointed to the old ones. > > Most probably these problems stem from an static errno. > > Another source can be a non reentrant gethostbyname. > > If dillo-0.7.3 work right on it, I'm puzzled. Today's cvs, run _locally_ on the ps2, looks fine so far (I _can_ induce a crash, by running remotely from _two_ machines at once; I recollect that when I got crashes a week or two ago I was running remotely from only one machine - (an old 486 running slack 4, which is libc5-based...) - but I can't reproduce that result ) Let's assume all is fixed unless I shout. Bob -- robert of northworthige

Jan. 31, 2004

Re: [Dillo-dev] building on AIX, gcc, dillo 0.7.3

by Jorge Arellano Cid

On Sat, 31 Jan 2004, Riccardo Mottola wrote: > on 1/28/04 5:58 PM, Andreas Schweitzer at > Andreas.Schweitzer(a)hs.uni-hamburg.de wrote: > > >>> > >>> and there I just found : > >>> > >>> http://aixpdslib.seas.ucla.edu/packages/gtk+.html > >> Unfortunately, aix 4.2 is quite out of date (but my original rs/6000 is > >> beginning to show its age too). > > > > I just checked. Back then on that machine I had/have : > > > > $ lslpp -l | egrep -i gtk > > freeware.gtk+.rte 1.2.8.0 COMMITTED Gimp ToolKit GUI Library > > freeware.gtk+.rte 1.2.8.0 COMMITTED Gimp ToolKit GUI Library > > > > So, I took it from one of these sites :-) > > > > Anyways, my hunch yesterday was right when I suspected a small patch > > for AIX. I have a diff for AIX lying around which is attached. > > > The fix produced a result: in fact build stops this way now: > gcc -DHAVE_CONFIG_H -I. -I. -I../.. -I/usr/local/include > -I/usr/local/includ > e/gtk-1.2 -I/usr/local/include/glib-1.2 -I/usr/local/lib/glib/include -g -O2 > -D_ > REENTRANT -D_THREAD_SAFE -Wall -Waggregate-return -c `test -f 'dpi.c' || > echo '. > /'`dpi.c > dpi.c: In function `Dpi_get_server_uds_name': > dpi.c:497: `PF_LOCAL' undeclared (first use in this function) > dpi.c:497: (Each undeclared identifier is reported only once > dpi.c:497: for each function it appears in.) > dpi.c: In function `a_Dpi_bye_dpid': > dpi.c:749: `PF_LOCAL' undeclared (first use in this function) > gmake[1]: *** [dpi.o] Error 1 There's no PF_LOCAL in the CVS tarball. Please try it. (http://www.dillo.org/test/dillo-0.8.0-pre.Jan30.tar.gz) Cheers Jorge.-

Jan. 31, 2004

Re: [Dillo-dev] building on AIX, gcc, dillo 0.7.3

by Riccardo Mottola

on 1/28/04 5:58 PM, Andreas Schweitzer at Andreas.Schweitzer(a)hs.uni-hamburg.de wrote: >>> >>> and there I just found : >>> >>> http://aixpdslib.seas.ucla.edu/packages/gtk+.html >> Unfortunately, aix 4.2 is quite out of date (but my original rs/6000 is >> beginning to show its age too). > > I just checked. Back then on that machine I had/have : > > $ lslpp -l | egrep -i gtk > freeware.gtk+.rte 1.2.8.0 COMMITTED Gimp ToolKit GUI Library > freeware.gtk+.rte 1.2.8.0 COMMITTED Gimp ToolKit GUI Library > > So, I took it from one of these sites :-) > > Anyways, my hunch yesterday was right when I suspected a small patch > for AIX. I have a diff for AIX lying around which is attached. > The fix produced a result: in fact build stops this way now: gcc -DHAVE_CONFIG_H -I. -I. -I../.. -I/usr/local/include -I/usr/local/includ e/gtk-1.2 -I/usr/local/include/glib-1.2 -I/usr/local/lib/glib/include -g -O2 -D_ REENTRANT -D_THREAD_SAFE -Wall -Waggregate-return -c `test -f 'dpi.c' || echo '. /'`dpi.c dpi.c: In function `Dpi_get_server_uds_name': dpi.c:497: `PF_LOCAL' undeclared (first use in this function) dpi.c:497: (Each undeclared identifier is reported only once dpi.c:497: for each function it appears in.) dpi.c: In function `a_Dpi_bye_dpid': dpi.c:749: `PF_LOCAL' undeclared (first use in this function) gmake[1]: *** [dpi.o] Error 1

Jan. 31, 2004

Re: [Dillo-dev] Automake and dynamic linking

by Riccardo

Jorge Arellano Cid wrote: > > Jorgen, > Andreas, > Riccardo, > Matti, > Bob > (portability issues here): > > Hi, here is a brief covering the latest portability patches now > in CVS. These are not final an certainly need testing, so lets > try to get it straight. > > Issues: libpng, pthreads, proper linking, AF_LOCAL, libc5, PS2 > I untarred and run configure, it configured to the end so it didn't stop on pthread or linpng. I didn't attempt to build since there is this GTK header problem. I try tp apply old Andreas' patch manually and check the effect. Bye

Jan. 31, 2004

Re: [Dillo-dev] Automake and dynamic linking

by Jorge Arellano Cid

Jorgen, Andreas, Riccardo, Matti, Bob (portability issues here): Hi, here is a brief covering the latest portability patches now in CVS. These are not final an certainly need testing, so lets try to get it straight. Issues: libpng, pthreads, proper linking, AF_LOCAL, libc5, PS2 libpng: ------ On Wed, 14 Jan 2004, Jorgen Viksell wrote: > Hi! > > tis 2004-01-06 klockan 16.01 skrev Jorge Arellano Cid: > > [...] > > We can't rely on libpng-config because it would break support > > for older distros (for instance slackware 3.9), and there're > > plenty of dillo users with old distros on laptops. > > Fair enough. Using Debian unstable I tend to forget about those people > stuck with non-evolving distros. > > > The current png detection routine has come a long way to be > > able to detect the library on a wide range of SW/HW combinations. > > Maybe the best approach is to try libpng-config first, and when > > not present, try the older libs. > > I made it do so. Could you please try out the fallback code, using: > PNG_CONFIG=foobar ./configure > And if it still fails could you send me the config.log file. Sorry for taking a long time to answer, I got a lot of troubles installing an old libc5 distro (Slackware 3.6) for testing. These were unrelated to the distro and not worth to detail here! ;) The simple fallback-test passed on my slk current, but the old installation payed off quickly, because it failed on it. For instance: <q> checking for libpng-config... missing checking png.h usability... yes checking png.h presence... yes checking for png.h... yes checking libpng/png.h usability... yes checking libpng/png.h presence... yes checking for libpng/png.h... yes checking for png_check_sig in -lpng... no configure: WARNING: *** No libpng found. Disabling PNG images *** </q> Fixed, added "-lm" and a "&& break" to avoid the second header test when the first is present. The version in CVS works both with the current and libc5 distros. As for the image disabling, although I don't see much use for it, I left to avoid some work at this stage. The one thing I don't like is the #ifdef stuff. Well, if it's not worth it can be removed. pthreads: -------- I added the "pthreads" (with an 's') to the search list as Andreas suggested. Riccardo: please test if this works for you. I made a tarball so you don't need to play with autoconf/automake/autogen. URL: http://www.dillo.org/test/dillo-0.8.0-pre.Jan30.tar.gz Note: there's still some work to do with this one, read below. proper linking: -------------- Now the linked libraries are ok (the problem was that dpis and dpid got all of the libraries used by dillo linked in). ...except for pthread that is till present on dpi, dpis. AF_LOCAL: -------- After studying the history of the socket library development and socket.h for linux, BSD, HURD, etc, it seems that all these are (should be) the same: PF_UNIX : old BSD name for PF_LOCAL PF_LOCAL : POSIX for AF_LOCAL PF_FILE : seldom used POSIX for AF_UNIX AF_LOCAL : somewhat old but the most compatible. AF_UNIX : oldest. Fixed this for the whole tarball. It works on both of my testing environments. Should work for everyone. libc5: ----- Now it compiles cleanly with libc5. Except for a warning on a repeated symbol between libjpeg and autoconf. --innocuous though PS2: --- > [Bob wrote] > - my experience of recent cvs, compiled on the playstion2, (blackrhino > 2.2.21 kernel), apparently shows a disconcerting feature. After a few > minutes of scrolling up and down standard web-sites the whole system > will hang (eg it's unpingable, and doesn't respond to any keyboard > input), and needs a hard reset :-( > > Anyone else currently testing on the ps2 please (and hopefully had > better experiences, or a fix)?? I don't have a PS2, but when testing on libc5, there was a similar problem (crash after scrolling, or crash instants after starting) when _some_ of the libraries used by the X server where compiled reentrant and some others not. In fact I was upgrading slk3.6's with the ones at: http://www.linuxsavvy.com/resources/ and some links pointed to the old ones. Most probably these problems stem from an static errno. Another source can be a non reentrant gethostbyname. If dillo-0.7.3 work right on it, I'm puzzled. That's it. This CVS should work on the same platforms than before and with a few more. Please test and report. Best Jorge.- PS: Those that have had problems with 0.7.3, please test this CVS version, or tarball: URL: http://www.dillo.org/test/dillo-0.8.0-pre.Jan30.tar.gz PS2: Ah there's also a fix to include "dpidc". It fixes "make dist".

Jan. 30, 2004