On 3/5/11 11:43 AM, Alan_H wrote:
Thanks to "vitamin" for the suggestion. I was trying to follow the regression instructions, and didn't realize "./configure" was part of the cycle after the first setup. However, to move on...
I went into the wine-git directory without having done anything since the last post, ran ./configure, then make depend, then make. All seemed to go well for the ./configue and make depend. This quote shows what happened during the make:
.... (many lines, a few hundred?, generated by the "make")
freetype.c:5031: error: called object ‘pFT_MulFix’ is not a function
freetype.c:5032: error: called object ‘pFT_MulFix’ is not a function
make[2]: *** [freetype.o] Error 1
make[2]: Leaving directory `/home/alan/wine-git/dlls/gdi32'
make[1]: *** [gdi32] Error 2
make[1]: Leaving directory `/home/alan/wine-git/dlls'
make: *** [dlls] Error 2
<prompt>
Two problems now:
1) to be clear, is this the cycle?
initial setup (downloads, compile, install problem application)
set up bisect (start, good, bad)
start of cycle
run the program
run bisect good /bad
exit cycle if converged
./configure
make depend
make
go to start of cycle
report bug with final data
This is my current understanding. Before I wasn't running "./configure" and "make depend" in the cycle
2) what do I have to do to make it possible to clear the present error / initial state so I can start the regression analysis again?
First, there is no need to do an initial build cycle. You can do this:
git clone...
git bisect start
git bisect good <tag>
git bisect bad <tag>
./configure (additional configuration tags go here)
make
Then run wine from the directory where you built wine.
If the program runs, the next step is to mark the bisect good. If the
program does not run, mark the bisect as bad. Repeat until the entire
process is done. Then you have to do ONE MORE run of bisect. You will
receive a message that the first bad commit is ..... You can remove the
commit and then try another build. It is interesting that some problems
have multiple origins based on several code additions/changes/deletions.
Did you read the Regression Testing page on the Wiki. It says what I
just said above.
James McKenzie
Thanks in advance for continuing support...
Alan
