Hi Yaacov, Thanks a lot for your build report - at least we know that there is hope in sight for us in Debian :) Am I right in guessing that you built OO using a custom built gcc 3.04 instead of the one in Debian? The impression I get from the guys on #Openoffice.org is that they usually use their own compiler (configure --with-gcc-home) so using a standard installed compiler is not so well tested, and that is one of the things that gives us headaches trying to build proper .debs :( I have been through your patches and investigated where you got them from. For the benefit of Jan and the rest of the list, I'll summarise my findings: On Mon, Apr 08, 2002 at 10:10:02AM +0300, Yaacov Akiba Slama wrote: > I succesfully built a working oo 641d in testing + unstable stuff with > gcc 3.04. I already mentioned gcc. Yaacov, what bison version did you use, BTW? We are working with OOO_STABLE_1, which is basically 641d+bugfixes, so most issues should be the same as Yaacov. > 1) Patch /usr/include/X11/Xlib.h to add a missing prototype: > > 2) Patch two files in the oo sources (ttimes.cxx from the last cvs of > oo, and thints.cxx to add the keyword 'class' 3 times). > > I didn't invent anything, I only saw these stuff in oo mailing lists. > > Patches enclosed > --- /usr/include/X11/Xlib.h.orig Sun Apr 7 02:02:10 2002 > +++ /usr/include/X11/Xlib.h Sun Apr 7 02:02:28 2002 > @@ -4615,6 +4615,12 @@ > #endif > ); > > +extern char *XSetIMValues( > +#if NeedVarargsPrototypes > + XIM /* im */, ... > +#endif > +); > + > extern Display *XDisplayOfIM( > #if NeedFunctionPrototypes > XIM /* im */ The Xlib.h patch came from the thread 'build experiences' by Noll Janos on March 4th: ------------------------- 2. XSetIMValues function definition missing This seems to be a problem by the XFree86 team. The "XSetIMValues" seems to mysteriously "fade out" from the header file, from time to time. To /usr/X11/include/X11/Xlib.h , i added: extern char *XSetIMValues( #if NeedVarargsPrototypes XIM /* im */, ... #endif ); -------------------------- I haven't seen this problem yet; it's not in the OOo package build. > --- sw/source/core/txtnode/thints.cxx.orig Sun Apr 7 22:51:22 2002 > +++ sw/source/core/txtnode/thints.cxx Sun Apr 7 22:55:13 2002 > @@ -1851,7 +1851,7 @@ > ( SFX_ITEM_SET == rNode.GetpSwAttrSet()->GetItemState( nWhich, > FALSE, &pParaItem ) ) && ( pParaItem == &pHint->GetAttr() ); > BOOL bReplace = !( SETATTR_DONTREPLACE & nMode ); > - SwpHtStart_SAR *pTmpHints = 0; > + class SwpHtStart_SAR *pTmpHints = 0; > > USHORT i; > // Wir wollen zwar von nHtStart bis nMaxEnd, muessen aber ggf. > @@ -2027,7 +2027,7 @@ > // Beim Einfuegen spaltet es sich selbst. > if( pHistory ) pHistory->Add( pOther ); > if( !pTmpHints ) > - pTmpHints = new SwpHtStart_SAR(); > + pTmpHints = new class SwpHtStart_SAR(); > pTmpHints->C40_INSERT( SwTxtAttr, pOther, > pTmpHints->Count() ); > Cut( i ); > @@ -2076,7 +2076,7 @@ > // Beim Einfuegen spaltet es sich selbst. > if( pHistory ) pHistory->Add( pOther ); > if( !pTmpHints ) > - pTmpHints = new SwpHtStart_SAR(); > + pTmpHints = new class SwpHtStart_SAR(); > pTmpHints->C40_INSERT( SwTxtAttr, pOther, > pTmpHints->Count() ); > Cut( i ); Again, the thints.cxx patch came from the same thread: -------------------------- 4. Error: in sw/source/core/txtnode/thints.cxx class 'SwpHtStart_SAR' is inaccessible I had to change 'SwpHtStart_SAR *pTmpHints = 0;' line to 'class SwpHtStart_SAR *pTmpHints = 0;', and similarly, had to add the 'class' keyword in two other places (why? gcc-3 problem?) -------------------------- Again, I haven't come across this yet and it's not in the OOo package build. > diff -u -r1.2 -r1.2.10.1 > --- tools/source/datetime/ttime.cxx 2001/06/11 17:17:46 1.2 > +++ tools/source/datetime/ttime.cxx 2002/03/19 13:38:21 1.2.10.1 I have had this problem with ttime.cxx too and the patch is already in debian/patches/008_clk_tck_gcc-3.diff So, what about the other two problems? It may be that they are already fixed in OOO_STABLE_1 so I won't blindly add the patches unless someone confirms that the problems are still present, but hopefully the comments above will mean that if we do see the problems they will be relatively easy to fix. I say relatively, because we can't go round patching xlib.h ourselves so we'll have to talk to Branden, and probably devise a workaround in the meantime. Meanwhile, Jan has been working on an stlport package that has been compiled with gcc-3 and I am doing occasional builds (not yet complete). Chris
Attachment:
pgp1q5et2mA1J.pgp
Description: PGP signature