On Thu, Oct 2, 2008 at 3:25 PM, James Hawkins jhawkins@codeweavers.com wrote:
Hi, I can't repro this problem on my machine, so I'm sending this patch to patchwatcher by way of wine-patches to get some info. Please do not apply.
Regarding our wineconf discussion about patchwatcher spam on wine-patches and filtering. Perhaps we need 2 email address patches can be sent to. My thinking is the flow could be something like:
1. By default everyone sends a patch to wine-patches 2. If a patch gets rejected the author gets a failure mail in private, and now retries or testing patches should go to wine-patches-testing N times until they get it right 3. once its right, patchwatch could then forward the new good patch to wine-patches or could send an email to the dev saying that its good enough to now send to wine-patches.
Doing it like this, if everyone follows the rules, we won't get a ton of patchwatcher retries filling up wine-patches.
On Thu, Oct 2, 2008 at 3:27 PM, Steven Edwards winehacker@gmail.com wrote:
On Thu, Oct 2, 2008 at 3:25 PM, James Hawkins jhawkins@codeweavers.com wrote:
Hi, I can't repro this problem on my machine, so I'm sending this patch to patchwatcher by way of wine-patches to get some info. Please do not apply.
Regarding our wineconf discussion about patchwatcher spam on wine-patches and filtering. Perhaps we need 2 email address patches can be sent to. My thinking is the flow could be something like:
- By default everyone sends a patch to wine-patches
- If a patch gets rejected the author gets a failure mail in private,
and now retries or testing patches should go to wine-patches-testing N times until they get it right 3. once its right, patchwatch could then forward the new good patch to wine-patches or could send an email to the dev saying that its good enough to now send to wine-patches.
Doing it like this, if everyone follows the rules, we won't get a ton of patchwatcher retries filling up wine-patches.
Just so everyone is clear, I emailed Dan before submitting the patches asking if there was a way to send a patch directly to patchwatcher and bypass wine-patches, but he said there's not a way yet and that sending the patch to wine-patches should be OK for now.
On Thu, Oct 2, 2008 at 3:27 PM, Steven Edwards winehacker@gmail.com wrote:
On Thu, Oct 2, 2008 at 3:25 PM, James Hawkins jhawkins@codeweavers.com wrote:
Hi, I can't repro this problem on my machine, so I'm sending this patch to patchwatcher by way of wine-patches to get some info. Please do not apply.
Regarding our wineconf discussion about patchwatcher spam on wine-patches and filtering. Perhaps we need 2 email address patches can be sent to. My thinking is the flow could be something like:
- By default everyone sends a patch to wine-patches
- If a patch gets rejected the author gets a failure mail in private,
and now retries or testing patches should go to wine-patches-testing N times until they get it right 3. once its right, patchwatch could then forward the new good patch to wine-patches or could send an email to the dev saying that its good enough to now send to wine-patches.
You'd still be sending the patch twice to wine-patches, and by that logic, for _EVERY_ patch, not just ones that were rejected.
I think the idea at Wineconf was the best: Submit patches to wine-patches Patchwatcher does its voodoo magic If patch is good -> wine-patches-approved If patch fails -> wine-patches-rejected + e-mail to author
However, I think your idea has merit...if Dan would be willing, could add a test e-mail, that doesn't spam everyone, for testing only on patchwater. I.e., patch-test@kegel.com. Results would then join the rest on the results page (or possibly just e-mailed back, but that seems like more work for little gain.
-Austin
On Thu, Oct 2, 2008 at 4:42 PM, Austin English austinenglish@gmail.com wrote:
You'd still be sending the patch twice to wine-patches, and by that logic, for _EVERY_ patch, not just ones that were rejected.
No. you would only have to send it again if the first one was rejected by patchwatcher. By default no ones work flow would change, just the email you are sending it to for retries saving everyones inbox on spam. If your patch was bad you would still have to do extra leg work but all of that would be done in private. If we get make test working in more places then its possible the number of first time failures to wine-patches will go down also if we can encourage people to run make test before submitting patches.
On Thu, Oct 2, 2008 at 3:47 PM, Steven Edwards winehacker@gmail.com wrote:
On Thu, Oct 2, 2008 at 4:42 PM, Austin English austinenglish@gmail.com wrote:
You'd still be sending the patch twice to wine-patches, and by that logic, for _EVERY_ patch, not just ones that were rejected.
No. you would only have to send it again if the first one was rejected by patchwatcher. By default no ones work flow would change, just the email you are sending it to for retries saving everyones inbox on spam. If your patch was bad you would still have to do extra leg work but all of that would be done in private. If we get make test working in more places then its possible the number of first time failures to wine-patches will go down also if we can encourage people to run make test before submitting patches.
-- Steven Edwards
"There is one thing stronger than all the armies in the world, and that is an idea whose time has come." - Victor Hugo
Here's how I understand your idea, correct me if I'm wrong:
WinedevA submits a patch to wine-patches. Crashes on patchwatcher, etc. WinedevA gets a private e-mail from patchwatcher saying "you're a failure, learn to code better, fix this, etc.". WinedevA fixes patch up, resubmits to wine-patches-testing multiple times, finally gets it right. He does a little dance, celebrates, etc. Patchwatcher then forwards this good patch to wine-patches.
WinedevB submits a patch to wine-patches, works fine. Passes patchwatcher, no valgrind warnings, etc. Because it passed, it then forwards it to wine-patches. Unless patchwatcher somehow keeps track of all patches that have failed and been resubmitted, which seems like extra work for nothing.
I thought we had a consensus on the process I mentioned earlier...
-Austin
2008/10/2 Steven Edwards winehacker@gmail.com:
On Thu, Oct 2, 2008 at 3:25 PM, James Hawkins jhawkins@codeweavers.com wrote:
Hi, I can't repro this problem on my machine, so I'm sending this patch to patchwatcher by way of wine-patches to get some info. Please do not apply.
Regarding our wineconf discussion about patchwatcher spam on wine-patches and filtering. Perhaps we need 2 email address patches can be sent to. My thinking is the flow could be something like:
- By default everyone sends a patch to wine-patches
- If a patch gets rejected the author gets a failure mail in private,
and now retries or testing patches should go to wine-patches-testing N times until they get it right 3. once its right, patchwatch could then forward the new good patch to wine-patches or could send an email to the dev saying that its good enough to now send to wine-patches.
Doing it like this, if everyone follows the rules, we won't get a ton of patchwatcher retries filling up wine-patches.
I don't like this idea because of the expression "If you gave an infinite number of monkeys an infinite number of typewriters they would eventually produce the complete works of Shakespeare." In other words, if someone were to just send patches all day to patchwatcher, changing one thing at a time until they passed then one would eventually get through. However, the result in this case would probably not be a quality patch since patchwatcher would not be able to catch all possible bugs.
We want to be able to track how many retries a patch has had to go through before getting through since the more it had the more bugs it is likely to have. Therefore, we want all patchwatcher failures to go to a separate mailing list where we can see if a patch is being sent repeatedly so that we can email them to say "hey, you may want to take a while to rethink this patch or ask for help on wine-devel".
James, please note that I'm not accusing you of being one of said monkeys, I realise this is an exception :-)
On Thu, Oct 2, 2008 at 5:21 PM, Rob Shearman robertshearman@gmail.com wrote:
I don't like this idea because of the expression "If you gave an infinite number of monkeys an infinite number of typewriters they would eventually produce the complete works of Shakespeare." In other words, if someone were to just send patches all day to patchwatcher, changing one thing at a time until they passed then one would eventually get through. However, the result in this case would probably not be a quality patch since patchwatcher would not be able to catch all possible bugs.
We want to be able to track how many retries a patch has had to go through before getting through since the more it had the more bugs it is likely to have. Therefore, we want all patchwatcher failures to go to a separate mailing list where we can see if a patch is being sent repeatedly so that we can email them to say "hey, you may want to take a while to rethink this patch or ask for help on wine-devel".
I agree with that. There is no reason why the wine-patches-testing thing could not be a separate mailing list that we could monitor as well. I just don't think it would need that high level of attention like wine-patches because it would be the exception case rather than the rule (assuming we get winetest passing for most people before they submit patches), interested parties could review it, albeit less often and still have the effect of filtering the spam from the main mailing list.
James, please note that I'm not accusing you of being one of said monkeys, I realise this is an exception :-)
Me either. =) It was just seeing this thread made me think about it more and it did not seem that we had clear consensus on this issue at wineconf. In the end it does not matter, Alexandre will decide what works best for him. I think thats going to end up being some sort of tagging system like we discussed so that all patches good and bad get through with review. It would just be nice to have an outside way to privately send patches to patchwatcher without spamming everyone else.
-
Austin English -
James Hawkins -
Rob Shearman -
Steven Edwards