On Tue, Jun 9, 2009 at 7:14 AM, Ben Kleinshacklein@gmail.com wrote:
2009/6/9 Uwe Bonnes bon@elektron.ikp.physik.tu-darmstadt.de:
> "Ben" == Ben Klein shacklein@gmail.com writes:
Ben> 2009/6/9 Uwe Bonnes bon@elektron.ikp.physik.tu-darmstadt.de: >> Hello, >> >> on a fresh tree check out and compiled successfully yesterday >> (090608) and updated today(090609), compile fails with: >> >> make[1]: Entering directory >> `/media/sda8/sdb8/spare/bon/Wine/wine-git/include' ../tools/widl/widl >> -I. -I. -I../include -I../include \ -h -H activaut.h activaut.idl >> ../tools/widl/widl -I. -I. -I../include -I../include \ -h -H >> activdbg.h activdbg.idl ../tools/widl/widl -I. -I. -I../include >> -I../include \ -h -H activscp.h activscp.idl oaidl.idl:121: error: >> parameter 'pvData' of (null) 'tagSAFEARRAY' \ cannot derive from void >> * make[1]: *** [activscp.h] Fehler 1 make[1]: Leaving directory >> `/media/sda8/sdb8/spare/bon/Wine/wine-git/include' >> >> This is on Suse 11.1, x86-64. I have not seens this reported >> before. Any hints for fixing (perhaps my setup)?
Ben> Please provide the git revision number, not just the date.
Trying to run "git bisect" and going back to de945384a4c1f593820e91811c1c04ff0263ca48 (the number gitk reports in the topmost entry) now everything compiled fine. Strange...
I've occasionally got compilation errors while doing a git bisect. As there is no way to fix this (assuming there's nothing wrong with libs on my system, since every other revision I try gets built fine) without manually hacking at the git repository (which is a very bad idea for all sorts of reasons), I just mark that revision as bad and continue.
That's a very poor practice, leading to wrong regression test results. The proper thing to do is either: A) fix the compile manually, then revert your change before the next git bisect {good,bad} B) use 'git bisect skip'.