On 2018-03-05 16:25 -0300, Herbert Fortes wrote: > Em 05-03-2018 15:45, Sven Joachim escreveu: >> Source: libcdk5 >> Version: 5.0.20161210-5 >> Severity: minor >> Tags: patch >> >> One symbol is missing from debian/libcdk5.symbols, leading to a warning >> from dpkg-gensymbols seen in the build logs. According to the CHANGES >> file it was added on 2016/11/19, so I have chosen 5.0.20161120 as the >> minimal version, as for the other symbols. See the attached patch. >> > > The .symbols file, to pass on all architectures, should not have all symbols.
Well, it does not _need_ to have all symbols, unless you pass -c2 or higher to dpkg-gensymbols. But I think -c4 is actually recommended to catch any problems that might result in an unexpected ABI change. > Just the common symbols for every arch. That's what I learned. > > Are you saying something different from that? Can you explain a bit more? I was referring to the following message from dpkg-gensymbols[1]: ,---- | dpkg-gensymbols: warning: some new symbols appeared in the symbols file: see diff output below | dpkg-gensymbols: warning: debian/libcdk5/DEBIAN/symbols doesn't match completely debian/libcdk5.symbols | --- debian/libcdk5.symbols (libcdk5_5.0.20161210_hurd-i386) | +++ dpkg-gensymbols0_FMk7 2018-01-27 17:37:28.000000000 +0000 | @@ -306,6 +306,7 @@ | positionCDKButton@Base 5.0.20161120 | positionCDKObject@Base 5.0.20161120 | raiseCDKObject@Base 5.0.20161120 | + reRegisterCDKObject@Base 5.0.20161210 | readFile@Base 5.0.20161120 | refreshCDKScreen@Base 5.0.20161120 | refreshCDKWindow@Base 5.0.20161120 `---- True, I have not checked all architectures, only a couple of them. But I don't think that libcdk5's symbols are arch-dependent at all. Cheers, Sven 1. https://buildd.debian.org/status/fetch.php?pkg=libcdk5&arch=hurd-i386&ver=5.0.20161210-5&stamp=1517074653&raw=0