RE: svn commit: r39926 - trunk
Bert Huijben <[email protected]>
| Newsgroups | gmane.comp.version-control.subversion.rapidsvn.devel,gmane.comp.version-control.subversion.svn |
|---|---|
| Message-ID | <011c01ca49ce$c4e25ba0$4ea712e0$__37793.500084854$1255195759$gmane$org@nl> |
> -----Original Message----- > From: Daniel Shahaf [mailto:[email protected]] > Sent: zaterdag 10 oktober 2009 19:15 > To: Bert Huijben > Cc: dev-lmwclWVctOZK/[email protected]; svn-lmwclWVctOZK/[email protected] > Subject: RE: svn commit: r39926 - trunk > > Bert Huijben wrote on Sat, 10 Oct 2009 at 18:51 +0200: > > > Fix the Windows build. > > > > > > * build.conf (libsvn_subr): Export svn_debug.h. > > > > This fixes the Windows shared library build for debug mode, but it > breaks > > the Windows build for release mode.. See the buildbots. > > > > Thanks, I haven't noticed they broke. > > > libsvn_subr.def : error LNK2001: unresolved external symbol > > svn_dbg__preamble > > libsvn_subr.def : error LNK2001: unresolved external symbol > svn_dbg__printf > > ..\..\..\Release\subversion\libsvn_subr\libsvn_subr-1.lib : fatal > error > > LNK1120: 2 unresolved externals > > > > What do you suggest then? Without this change I get the same error > when > I try to use SVN_DBG in debug builds. > > Could we just define the svn_dbg__* functions unconditionally? (and > make them no-ops if SVN_DEBUG is not defined) I think that would be safe. But we might just define them always as the macros remove the call sites anyway. (And this would allow using them from a debug mode application compiled against a release library). Bert ------------------------------------------------------ http://subversion.tigris.org/ds/viewMessage.do?dsForumId=495&dsMessageId=2406016