On 2017-01-18 18:54, Emilio Pozuelo Monfort wrote: > On 18/01/17 17:38, Andreas Beckmann wrote: >> Package: release.debian.org
>> nmu madplay_0.15.2b-8.1 . ANY . unstable . -m "Rebuild against libid3tag >> 0.15.1b-12" >> >> adequate reports: >> >> madplay: symbol-size-mismatch /usr/bin/madplay => id3_build >> >> madplay was built against libid3tag 0.15.1b-11 (which was itself built >> in 2014), while 0.15.1b-12 was built two years later with a newer >> toolchain ... > > If libid3tag's ABI changed, then it needs to rename it's binary package. It's a const char[] that changed content (and length) between builds, not sure whether this qualifies for an ABI break. But I agree that such symbols should not be used. char const id3_build[] = "" # if defined(DEBUG) "DEBUG " # elif defined(NDEBUG) "NDEBUG " # endif # if defined(EXPERIMENTAL) "EXPERIMENTAL " # endif ; > Also have you checked that no other rdeps are affected? It's the only package in sid and stretch where adequate (called via piuparts) reports this mismatch. in jessie (0.15.1b-11): $ nm -D -S /usr/lib/libid3tag.so.0.3.0 | grep id3_build 0000000000010490 0000000000000008 R id3_build $ hd -n 8 -s 0x10490 /usr/lib/libid3tag.so.0.3.0 00010490 4e 44 45 42 55 47 20 00 |NDEBUG .| in sid (0.15.1b-12): $ nm -D -S /usr/lib/x86_64-linux-gnu/libid3tag.so.0.3.0 | grep id3_build 0000000000010b10 0000000000000001 R id3_build $ hd -n 1 -s 0x10b10 /usr/lib/x86_64-linux-gnu/libid3tag.so.0.3.0 00010b10 00 |.| Andreas