Hello, as you might have already noticed I took the chance while upgrading SVN server to 1.4.0 and creating a mirror to effectively set it to readonly state for now.
The blocker bug is the stack corruption issue, which somewhen (though rather frequently, but to be fully sure one needs at least 10 boots) results in "IoPageWrite() failed" during setup.
Herve already took a lot of time to try to track down the revision number of the regression, however I'm not aware of his recent achievements.
I would like to ask everyone (especially those who wants to become testers :) ) to please help and find the revision number which has this regression. Further commits can go only when this is fixed (the concept of bootable trunk).
WBR, Aleksey Bragin.
Thanks to Herve, Quip, Filip and of course Alex the bugs I was talking about were fixed.
I promised for many days that I will post the rules we came up with WaxDragon and now the time came.
We are introducing a concept of a "working" trunk. It means, that every revision *must* boot to explorer (aka 3rd boot) and, doesn't crash from doing simple actions like clicking start button and double clicking my computer on the desktop.
To aid in this, an automatic regression testing system is going to be added to the existing buildbot, which will inform (via mailing list, I think or via IRC channel) if some tests fail (as it's supposed to do now when compiling breaks).
Before automatic regression system is set up, I strongly ask to
*not* *commit* *any* *new* *features* (unless they are disabled by default, like Alex's LPC changes guarded by NTLPC define) and commit only bugfixes which you tested at least 10 times.
This semi-freeze should not be long, considering automatic regression system already seems to work and we need to integrate it to buildbot and then start improving it more.
Thanks for understanding and for your patience related to recent switch of SVN to readonly mode. I will keep posting to this thread with regards to progress with the testing system / SVN readiness for the usual commit-flow.
With the best regards, Aleksey Bragin.
On Nov 4, 2006, at 1:11 PM, Aleksey Bragin wrote:
Hello, as you might have already noticed I took the chance while upgrading SVN server to 1.4.0 and creating a mirror to effectively set it to readonly state for now.
The blocker bug is the stack corruption issue, which somewhen (though rather frequently, but to be fully sure one needs at least 10 boots) results in "IoPageWrite() failed" during setup.
Herve already took a lot of time to try to track down the revision number of the regression, however I'm not aware of his recent achievements.
I would like to ask everyone (especially those who wants to become testers :) ) to please help and find the revision number which has this regression. Further commits can go only when this is fixed (the concept of bootable trunk).
WBR, Aleksey Bragin.
Update regarding current state of trunk and state of commits:
Bug with bootcd (IoPageWrite failed) has been fixed, and a number of other bugs too (Thank You, developers!). We are left with a LiveCD not working bug (again, the same IoPageWrite () failed, which probably means another stack corruption somewhere during boot sequence). I suppose this is the only major issue left in trunk. Also, GDB support is broken (/DEBUGPORT=GDB doesn't really switch GDB mode on, it just seems to be ignored).
Regarding codefreeze - new features can be committed now, but please test them very thoroughly, and for the time being (until automatic regression testing system is in place) I test trunk for brokenness (other are encouraged to help testing too). If trunk is broken, no other commits are allowed other than commit fixing it.
If you have changes, which lead to increasing trunk's instability, better create a branch and commit there, while debugging that code.
With the best regards, Aleksey Bragin.
On Nov 4, 2006, at 1:11 PM, Aleksey Bragin wrote:
Hello, as you might have already noticed I took the chance while upgrading SVN server to 1.4.0 and creating a mirror to effectively set it to readonly state for now.
The blocker bug is the stack corruption issue, which somewhen (though rather frequently, but to be fully sure one needs at least 10 boots) results in "IoPageWrite() failed" during setup.
Herve already took a lot of time to try to track down the revision number of the regression, however I'm not aware of his recent achievements.
I would like to ask everyone (especially those who wants to become testers :) ) to please help and find the revision number which has this regression. Further commits can go only when this is fixed (the concept of bootable trunk).
WBR, Aleksey Bragin.
-
Aleksey Bragin