Hi all. There seem to exist certain inconsistency between which compilers are marked as "required" in explicit-failures-markup.xml, and how often the regressions are run on these compilers. For example, for a long time there has been no results for msvc or vc-7_0. The cw-8_3 appeared in the regression tables some time ago (adding 20 or so failures to the typeof test results), had the regressions run a couple of times, and vanished (probably somebody removed the old results). This situation makes it very difficult to address the issues on these compilers. For example very few or no people on this list have cw-8_3 installed. The "batch mode" (modifying the library or tests and waiting for the next regression results) is almost not practical, since the regressions on this compiler are run once a week or even less often. Is there any other possibility that I am missing? At the same time at least three parties run the regressions on vc7_1. I wonder if such resource utilization is wise... I think regression testing is essential, and, ideally, should be run before every commit. I also understand that with Boost this is not feasible. But running regressions at least daily for every required compiler should be a must, IMO. If we can't achieve this for some compilers, this compilers should be, I think, downgraded to "optional". Regards, Arkadiy
are run on these compilers. For example, for a long time there has been no results for msvc or vc-7_0. [..]
This situation makes it very difficult to address the issues on these compilers.
I would like to reemphasize that unless somebody running regression tests for old compilers I have no way to make sure Boost.Test is working. Genandiy
Arkadiy Vertleyb writes:
There seem to exist certain inconsistency between which compilers are marked as "required" in explicit-failures-markup.xml, and how often the regressions are run on these compilers.
Some, yes, in particular because the list hasn't been updated yet for the upcoming release.
For example, for a long time there has been no results for msvc or vc-7_0.
They are back online now and will stay that way.
The cw-8_3 appeared in the regression tables some time ago (adding 20 or so failures to the typeof test results), had the regressions run a couple of times, and vanished (probably somebody removed the old results).
We were actually thinking about permanently dropping out cw-8_3 from our tests, but I guess following the principle of not dropping support without giving one release worth of warning means that we should stick with it for one more round. Hmm, David, is there a chance you could pick this one up from us? Alternatively, we can pick it up ourselves if somebody takes MSVC 6.5/7.0 from us (we would like to keep our regression cycle under 12 hours to provide a more or less adequate "feedback time"). -- Aleksey Gurtovoy MetaCommunications Engineering
Aleksey Gurtovoy <agurtovoy@meta-comm.com> writes:
Arkadiy Vertleyb writes:
There seem to exist certain inconsistency between which compilers are marked as "required" in explicit-failures-markup.xml, and how often the regressions are run on these compilers.
Some, yes, in particular because the list hasn't been updated yet for the upcoming release.
For example, for a long time there has been no results for msvc or vc-7_0.
They are back online now and will stay that way.
The cw-8_3 appeared in the regression tables some time ago (adding 20 or so failures to the typeof test results), had the regressions run a couple of times, and vanished (probably somebody removed the old results).
We were actually thinking about permanently dropping out cw-8_3 from our tests, but I guess following the principle of not dropping support without giving one release worth of warning means that we should stick with it for one more round.
Hmm, David, is there a chance you could pick this one up from us? Alternatively, we can pick it up ourselves if somebody takes MSVC 6.5/7.0 from us (we would like to keep our regression cycle under 12 hours to provide a more or less adequate "feedback time").
I could in theory do any of these. I will have to wipe a linux machine (I'm not using currently) and install Windows on it, though. VC7 got wiped off my disk and I think I don't have install disks anymore, so I have to get someone to send it to me. -- Dave Abrahams Boost Consulting www.boost-consulting.com
David Abrahams writes:
Aleksey Gurtovoy <agurtovoy@meta-comm.com> writes:
Hmm, David, is there a chance you could pick this one up from us? Alternatively, we can pick it up ourselves if somebody takes MSVC 6.5/7.0 from us (we would like to keep our regression cycle under 12 hours to provide a more or less adequate "feedback time").
I could in theory do any of these. I will have to wipe a linux machine (I'm not using currently) and install Windows on it, though. VC7 got wiped off my disk and I think I don't have install disks anymore, so I have to get someone to send it to me.
Even if you just take MSVC 6.5 or MWCW 8.3, that would shorten our cycle by ~2 hours and would be much appreciated. -- Aleksey Gurtovoy MetaCommunications Engineering
Aleksey Gurtovoy <agurtovoy@meta-comm.com> writes:
Hmm, David, is there a chance you could pick this one up from us? Alternatively, we can pick it up ourselves if somebody takes MSVC 6.5/7.0 from us (we would like to keep our regression cycle under 12 hours to provide a more or less adequate "feedback time").
There's a chance, but it will require me to wipe an unused Linux machine and install Windows there, with compilers. I'm not sure whether I'll have time before I leave at the end of this week (I'll be gone all next week). -- Dave Abrahams Boost Consulting www.boost-consulting.com
David Abrahams wrote:
Aleksey Gurtovoy <agurtovoy@meta-comm.com> writes:
Hmm, David, is there a chance you could pick this one up from us? Alternatively, we can pick it up ourselves if somebody takes MSVC 6.5/7.0 from us (we would like to keep our regression cycle under 12 hours to provide a more or less adequate "feedback time").
actually, I could take VC6.5 (also with Dinkumware 3.08) and *maybe* also vc7.0 B.
Bronek Kozicki writes:
David Abrahams wrote:
Aleksey Gurtovoy <agurtovoy@meta-comm.com> writes:
Hmm, David, is there a chance you could pick this one up from us? Alternatively, we can pick it up ourselves if somebody takes MSVC 6.5/7.0 from us (we would like to keep our regression cycle under 12 hours to provide a more or less adequate "feedback time").
actually, I could take VC6.5 (also with Dinkumware 3.08) and *maybe* also vc7.0
That would be great! And thanks! -- Aleksey Gurtovoy MetaCommunications Engineering
David Abrahams writes:
There's a chance, but it will require me to wipe an unused Linux machine and install Windows there, with compilers. I'm not sure whether I'll have time before I leave at the end of this week (I'll be gone all next week).
Understood. We will take Bronek's offer WRT MSVC 6.5, but if you still want to take on CW 8.3/VC 7.0 (however late), we will only be grateful. -- Aleksey Gurtovoy MetaCommunications Engineering
Aleksey Gurtovoy <agurtovoy@meta-comm.com> writes:
David Abrahams writes:
There's a chance, but it will require me to wipe an unused Linux machine and install Windows there, with compilers. I'm not sure whether I'll have time before I leave at the end of this week (I'll be gone all next week).
Understood. We will take Bronek's offer WRT MSVC 6.5, but if you still want to take on CW 8.3/VC 7.0 (however late), we will only be grateful.
I'll try to make it happen. No promises, though (sorry). -- Dave Abrahams Boost Consulting www.boost-consulting.com
participants (5)
-
Aleksey Gurtovoy -
Arkadiy Vertleyb -
Bronek Kozicki -
David Abrahams -
Gennadiy Rozental