Regression test - bad rev input

[Col]

Hi,
I have been trying to do a regression test between 1.1.10 and 1.1.12, as there seems to be a number of problems with the installation of the Macromedia suite (Dreamweaver, Fireworks, Flash).
I think I have ccache installed.
I think I have git OK.
I run git bisect start ok,
Then when I run git bisect good wine 1.1.10, I get:
Bad rev input: wine 1.1.10.

What am I doing wrong?

[Col]

I have found the problem,
I mistyped the line.
Should have been
git bisect good wine-1.1.10
(Missed the -hyphen)

[Col]

Now when I did:

Code: Select all

git bisect bad wine-1.1.12

I got:

Code: Select all

Bisecting: 467 revisions left to test after this
fatal: Entry '.gitignore' not uptodate. Cannot merge.

This does not look good.
I have checked my typing this time.
I must be doing something wrong.

[austin987]

On Thu, Jan 8, 2009 at 1:47 PM, Col <[email protected]> wrote:

I have found the problem,
I mistyped the line.
Should have been
git bisect good wine-1.1.10
(Missed the -hyphen)

Upgrade to git, the problem's likely fixed.

--
-Austin

[austin987]

On Thu, Jan 8, 2009 at 1:55 PM, Col <[email protected]> wrote:

Now when I did:

Code:
git bisect bad wine-1.1.12


I got:

Code:
Bisecting: 467 revisions left to test after this
fatal: Entry '.gitignore' not uptodate. Cannot merge.


This does not look good.
I have checked my typing this time.
I must be doing something wrong.

Try:

$ git bisect reset
$ git reset --hard origin
$ git fetch
$ git rebase origin

That will get you the current git, try building that and see if it works.


--
-Austin

[Col]

Austin, Thanks.
I have done

Code: Select all

 sudo apt-get install git-core

and it said:

Code: Select all

Reading package lists... Done
Building dependency tree       
Reading state information... Done
git-core is already the newest version.
0 upgraded, 0 newly installed, 0 to remove and 0 not upgraded.

Also, it shows I already have version 1.5.4.3
Now when I run:

Code: Select all

git clone git://source.winehq.org/git/wine.git wine-git

I get:

Code: Select all

'wine-git' already exists.

[vitamin]

Now when I run:

Code: Select all

git clone git://source.winehq.org/git/wine.git wine-git

I get:

Code: Select all

'wine-git' already exists.

Don't do that. Pull the latest GIT changes you need to:

Code: Select all

cd wine-git
git pull

[Col]

Thanks Vitamin.
I have now done the

Code: Select all

cd wine-git
git pull

Seemed to work except for the last line which said:

Code: Select all

fatal: Entry '.gitignore' not uptodate. Cannot merge.

So which step of regression testing should I be up to now?

[Col]

Code: Select all

$ git bisect reset
$ git reset --hard origin
$ git fetch
$ git rebase origin

Now done that Austin - thanks,
Now I seem to be on my way...

[Col]

All was going fine with this regression, twice through.
Then when i got to the time when I enter:

Code: Select all

CC="ccache gcc" ./configure --verbose && make depend && make

The last few lines of output are:

Code: Select all

ccache gcc -c -I. -I. -I../../include -I../../include  -D__WINESRC__  -D_REENTRANT -fPIC -Wall -pipe -fno-strict-aliasing -Wdeclaration-after-statement -Wwrite-strings -Wpointer-arith  -g -O2  -o schannel.o schannel.c
schannel.c:61: error: ‘gnutls_mac_get_key_size’ undeclared here (not in a function)
schannel.c:61: warning: type defaults to ‘int’ in declaration of ‘pgnutls_mac_get_key_size’
schannel.c: In function ‘schan_QueryContextAttributesW’:
schannel.c:856: error: called object ‘pgnutls_mac_get_key_size’ is not a function
make[2]: *** [schannel.o] Error 1
make[2]: Leaving directory `/home/colin/wine-git/dlls/secur32'
make[1]: *** [secur32] Error 2
make[1]: Leaving directory `/home/colin/wine-git/dlls'
make: *** [dlls] Error 2

This does not look good.
What should I do?

[Col]

Thought I would add that I have checked for dependencies using the list provided on the Regression testing page, and the list indicates that they are already the newest version for each.

If I try to keep going, when I get to install the application I get:

Code: Select all

wine: created the configuration directory '/home/colin/.wine'
wine client error:9: version mismatch 352/357.
Your wineserver binary was not upgraded correctly,
or you have an older one somewhere in your PATH.
Or maybe the wrong wineserver is still running?

Should I just reset the bisect and start again, or will the same thing happen?

[vitamin]

wine client error:9: version mismatch 352/357.

The older/newer version of wineserver is still running. Make sure all parts of Wine stop before you test next compile.

[Vincent Povirk]

The last few lines of output are:

Code:
ccache gcc -c -I. -I. -I../../include -I../../include -D__WINESRC__ -D_REENTRANT -fPIC -Wall -pipe -fno-strict-aliasing -Wdeclaration-after-statement -Wwrite-strings -Wpointer-arith -g -O2 -o schannel.o schannel.c
schannel.c:61: error: 'gnutls_mac_get_key_size' undeclared here (not in a function)
schannel.c:61: warning: type defaults to 'int' in declaration of 'pgnutls_mac_get_key_size'
schannel.c: In function 'schan_QueryContextAttributesW':
schannel.c:856: error: called object 'pgnutls_mac_get_key_size' is not a function
make[2]: *** [schannel.o] Error 1
make[2]: Leaving directory `/home/colin/wine-git/dlls/secur32'
make[1]: *** [secur32] Error 2
make[1]: Leaving directory `/home/colin/wine-git/dlls'
make: *** [dlls] Error 2

Sounds like that revision of Wine had a compilation bug.

Have you compiled and tested any versions of Wine earlier than this?

If an older version compiles, this bug was a regression, and there is
a range of broken commits that you just happened to hit. If that
happened, type "git bisect skip" to mark this revision as an
untestable one. Then type "git bisect visualize" to see a list of
commits left to test, and select one far from the middle or either
extreme. In the SHA1 ID field, you'll see a hex string something like
"c8cf4211869ea3de5bbc9b06c23e8b37ce27f319". Write down the first 5 or
6 digits of the id, close the gitk window, and check out that commit
by typing "git checkout c8cf42". Then compile and test as usual.

If you can't compile any versions older than that one (or if too many
revisions are broken), you'll need to follow the instructions on this
page: http://wiki.winehq.org/ReverseRegressionTesting

[austin987]

On Fri, Jan 9, 2009 at 11:23 AM, Vincent Povirk
<[email protected]> wrote:

The last few lines of output are:

Code:
ccache gcc -c -I. -I. -I../../include -I../../include -D__WINESRC__ -D_REENTRANT -fPIC -Wall -pipe -fno-strict-aliasing -Wdeclaration-after-statement -Wwrite-strings -Wpointer-arith -g -O2 -o schannel.o schannel.c
schannel.c:61: error: 'gnutls_mac_get_key_size' undeclared here (not in a function)
schannel.c:61: warning: type defaults to 'int' in declaration of 'pgnutls_mac_get_key_size'
schannel.c: In function 'schan_QueryContextAttributesW':
schannel.c:856: error: called object 'pgnutls_mac_get_key_size' is not a function
make[2]: *** [schannel.o] Error 1
make[2]: Leaving directory `/home/colin/wine-git/dlls/secur32'
make[1]: *** [secur32] Error 2
make[1]: Leaving directory `/home/colin/wine-git/dlls'
make: *** [dlls] Error 2

Sounds like that revision of Wine had a compilation bug.

Have you compiled and tested any versions of Wine earlier than this?

If an older version compiles, this bug was a regression, and there is
a range of broken commits that you just happened to hit. If that
happened, type "git bisect skip" to mark this revision as an
untestable one. Then type "git bisect visualize" to see a list of
commits left to test, and select one far from the middle or either
extreme. In the SHA1 ID field, you'll see a hex string something like
"c8cf4211869ea3de5bbc9b06c23e8b37ce27f319". Write down the first 5 or
6 digits of the id, close the gitk window, and check out that commit
by typing "git checkout c8cf42". Then compile and test as usual.

If you can't compile any versions older than that one (or if too many
revisions are broken), you'll need to follow the instructions on this
page: http://wiki.winehq.org/ReverseRegressionTesting

8c4a2978b281e9519a6581d859add636de4b42d4 is the patch fixing this.

--
-Austin

[Col]

8c4a2978b281e9519a6581d859add636de4b42d4 is the patch fixing this.

Am I supposed to do anything with this piece of information?

I've done the git bisect skip. Doing the visualize step required an installation, but it is well worth it, as the visualisation makes me feel more comfortable seeing what is going on.
The one that I 've now done the skip on is:
947976f22cb80a0cb14cc3da71622a62f9628968

I have re-started with one on the good side, (b15ba). I think I am on my way again.

[Col]

Thanks Austin, I eventually worked out the significance of what you posted.
With the great pointers of using gitk, I used it to help me narrow it down. I have been able to say that the regression for installing Dreamweaver et al occurred between:
Bisect good:
3c114a88ca1752ac07fe039d0471f89384ea4704
and Bisect bad:
8c4a2978b281e9519a6581d859add636de4b42d4

As I cannot work out what to do between these, I feel I am stuck as I cannot test what I cannot compile. I can only post this information in the Bugzilla, though I am not sure how it will be useful, with 23 revisions yet to test.

[austin987]

On Fri, Jan 9, 2009 at 7:19 PM, Col <[email protected]> wrote:

Thanks Austin, I eventually worked out the significance of what you posted.
With the great pointers of using gitk, I used it to help me narrow it down. I have been able to say that the regression for installing Dreamweaver et al occurred between:
Bisect good:
3c114a88ca1752ac07fe039d0471f89384ea4704
and Bisect bad:
8c4a2978b281e9519a6581d859add636de4b42d4

As I cannot work out what to do between these, I feel I am stuck as I cannot test what I cannot compile. I can only post this information in the Bugzilla, though I am not sure how it will be useful, with 23 revisions yet to test.

Have you tried in current git? I may have missed you saying that...

--
-Austin

[Col]

Dear Austin,

I started with wine 1.1.12 as bad.
What do you mean by current git?

[austin987]

On Fri, Jan 9, 2009 at 7:33 PM, Col <[email protected]> wrote:

Dear Austin,

I started with wine 1.1.12 as bad.
What do you mean by current git?

$ git reset --hard origin
$ git fetch ; git rebase origin
$ ./configure && make depend && make

That will give you the current development version. Try that. 1.1.12
had a couple major regressions.

--
-Austin

[Col]

Thanks Austin,

I have now also tested on the current git.
Installation fails on the current git. (Current git is bad).

My son worked out from gitk what was involved in doing a bisect by taking the fix in 8c4a2978b281e9519a6581d859add636de4b42d4 and applying it where the bisect would have been done. This was done by manually editing the configure and configure.ac with the changes in the fix (after backing up from the changes), and compiling. (We reversed our manual changes after testing).

For example this showed that 1a71479fbde0c10ad946334c524c3fdcb7dfab01 is good.

Is there any problem with continuing with this method?

[austin987]

On Fri, Jan 9, 2009 at 9:32 PM, Col <[email protected]> wrote:

Thanks Austin,

I have now also tested on the current git.
Installation fails on the current git. (Current git is bad).

My son worked out from gitk what was involved in doing a bisect by taking the fix in 8c4a2978b281e9519a6581d859add636de4b42d4 and applying it where the bisect would have been done. This was done by manually editing the configure and configure.ac with the changes in the fix (after backing up from the changes), and compiling. (We reversed our manual changes after testing).

For example this showed that 1a71479fbde0c10ad946334c524c3fdcb7dfab01 is good.

Is there any problem with continuing with this method?

No, that should work fine. See http://wiki.winehq.org/ReverseRegressionTesting

--
-Austin