Re: CppCheck Report on vm140 Not Running
On Thu, 29 Apr 2021 at 18:26:27 +, Luke Benes wrote: > Whether Maaten misremembered, there was a net spit, or > problem with your search, he did reach out to the dev list: > http://document-foundation-mail-archive.969070.n3.nabble.com/Automated-cppcheck-reports-not-running-tt4290186.html > Which is not a channel to reach the infra team, see https://tdf.io/infra . I for one don't read that mailing list :-P It is, however, probably the best place to gauge developer interest in having that service (it's not up to the infra team to decide whether there in interest in this or that service). Can only assume that if the service was relied upon by many then someone would have forwarded the message; like bugzilla issues reported to the QA list, weblate issues reported to the l10n list, any issue reported to the localized lists, etc. >> I'm very much against re-hooking a box > > A box that ran for years without any issues. When Maartin informed > us that he lost his vm140 ssh-key in his very first email weeks ago, > you said all you could do is reboot it from the hypervisor. Please don't change the narrative. I didn't write that's all I *could* do, only that “I'm not taking care of that VM. Rebooted it again, but that's I'll do for now.” And the message you're referring to appears to be what prompted you to write to Maarten in the first place — it was sent *before* Maarten's reply. > Why that was fix acceptable at the time? How is having a vm with a > lost ssh key preferable to resetting it? I didn't realize that the box was actually unmaintained, my bad. That fact indeed came to light early in the other thread, and the box is shut down since then. > It’s strange that all these emails and 2 weeks later, this new excuse > comes out of nowhere. Out of nowhere? On April 15 already you got a message from Florian which says a fresh setup would likely be preferable. Followed by two parallels threads discussing the usefulness of that service, and my reading is that there is little interest in reviving the service. See yesterday's message from Florian for the summary. > We are volunteers. Poor communication and artificial roadblocks only > discourages outside help. I disagree about the “artificial” qualification. Sorry to hear that you'd prefer having that service live but again see Florian's summary, there doesn't appear to have enough interest *at this point in time* to justify the extra work. Compare that with a Wiki plugin that's only use by a handful of pages and that causes us headaches during each upgrade because it's not maintained upstream — at some point it's worth asking ourselves if we're not better off without it (been there). -- Guilhem. signature.asc Description: PGP signature ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: CppCheck Report on vm140 Not Running
Whether Maaten misremembered, there was a net spit, or problem with your search, he did reach out to the dev list: http://document-foundation-mail-archive.969070.n3.nabble.com/Automated-cppcheck-reports-not-running-tt4290186.html > I'm very much against re-hooking a box A box that ran for years without any issues. When Maartin informed us that he lost his vm140 ssh-key in his very first email weeks ago, you said all you could do is reboot it from the hypervisor. Why that was fix acceptable at the time? How is having a vm with a lost ssh key preferable to resetting it? I was told to file a Redmond ticket, where again, no one from infra mentioned this, only the issue with credentials. https://redmine.documentfoundation.org/issues/3520#change-21592 It’s strange that all these emails and 2 weeks later, this new excuse comes out of nowhere. We are volunteers. Poor communication and artificial roadblocks only discourages outside help. ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: CppCheck Report on vm140 Not Running
Hi Maarten, On Thu, 29 Apr 2021 at 19:37:45 +0200, Maarten Hoes wrote: > On Thu, Apr 29, 2021 at 7:30 PM Guilhem Moulin > wrote: > Thanks for searching through the logs to see if I indeed posted on IRC > about this. So it looks like my memory is failing me on this point. You might have posted about it, all I'm saying is that if you did then I didn't see it :-) -- Guilhem. ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: CppCheck Report on vm140 Not Running
Hi, On Thu, Apr 29, 2021 at 7:30 PM Guilhem Moulin wrote: > On Thu, 29 Apr 2021 at 15:53:46 +, Luke Benes wrote: > > At the time, Maarten reported this on the dev IRC, infra IRC, and the > > mailing list. At the time, I assumed someone from infra would take > > care of it. > > As cloph wrote: > > | But also don't expect Infra team to browse the list for random > | infra-requests. For larger stuff a redmine ticket is the best way, for > | smaller bits here and there #tdf-infra on IRC or mail to > | hostmas...@documentfoundation.org. > > Also FWIW I don't see any relevant occurrence of cppcheck/vm140 in my > #tdf-infra and #libreoffice-dev 2020/2021 IRC logs. Might have happen > during a netsplit/disconnect/vacation, but IRC workflow is not “fire and > forget”, if the message is not acked then don't assume it's been seen > let alone will be acted upon. > > Thanks for searching through the logs to see if I indeed posted on IRC about this. So it looks like my memory is failing me on this point. Perhaps the posts were about the lcov reports instead (which also stopped running quite some time ago), or I just never went through with reporting it at all and now my memory of it is that I did do it. Again, sorry for the unintended confusion this may have caused. - Maarten ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: CppCheck Report on vm140 Not Running
On Thu, 29 Apr 2021 at 15:53:46 +, Luke Benes wrote: > At the time, Maarten reported this on the dev IRC, infra IRC, and the > mailing list. At the time, I assumed someone from infra would take > care of it. As cloph wrote: | But also don't expect Infra team to browse the list for random | infra-requests. For larger stuff a redmine ticket is the best way, for | smaller bits here and there #tdf-infra on IRC or mail to | hostmas...@documentfoundation.org. Also FWIW I don't see any relevant occurrence of cppcheck/vm140 in my #tdf-infra and #libreoffice-dev 2020/2021 IRC logs. Might have happen during a netsplit/disconnect/vacation, but IRC workflow is not “fire and forget”, if the message is not acked then don't assume it's been seen let alone will be acted upon. > Why can't you reset the user's password? I'm very much against re-hooking a box that has not been maintained for years to the main infrastructure and pretend all is well. > In the time, it took you to write that email, I could have rebooted > into recovery mode, dropped to root shell, and reset the root password > on the boxes I manage. This is a 30-second procedure. Every > hypervisor has some method to allow admins to boot guests into single > user. It's never been about being unable to regain access… -- Guilhem. signature.asc Description: PGP signature ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: CppCheck Report on vm140 Not Running
Hi, I wasn't subscribed to the dev mailing list up until just now, so when people stopped explicitly CC'ing me I missed some parts of this discussion. So here are some of my thoughts on the matter. I personally feel that if there is little interest in running cppcheck on the entire codebase regularly, I see little point of reviving the job. If there is/would be enough interest, I am willing to give it a shot to see if I can revive the job either on the current vm or on a new vm. No guarantees, other than to give it an honest attempt though. To me, Cloph made a fair point of at least documenting any manual steps that need to be taken. There are indeed a few manual steps, and I agree these would need to be documented at least and preferably automated in some fashion. I can't recall all the steps right now, but I'm fairly sure it will come back if I would try to revive the job. (For example, I created an gmail account and subscribed it to the dev mailing list so job success/failure emails could be send to the list.) To me, Luke made a fair argument that it might not be 'just' the 'code development team' that might be interested; he git grepped some and it looks like gsoc students (or other causal committers) have been using the tool to find 'easy hacks' or similar. These people might not be permanently watching mailing list messages or random irc chats when someone asks if anyone is interested in the job. I am personally not interested in maintaining the job (or the vm it runs on) in the long run. I do feel that the discussion has become a little overheated at the moment though, so perhaps it would be best if I just back off now. - Maarten -- Sent from: http://document-foundation-mail-archive.969070.n3.nabble.com/Dev-f1639786.html ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: CppCheck Report on vm140 Not Running
Cloph, > But given that the job had stopped so many months ago without anyone > complaining At the time, Maarten reported this on the dev IRC, infra IRC, and the mailing list. At the time, I assumed someone from infra would take care of it. > Recreating it is always an option Why can't you reset the user's password? In the time, it took you to write that email, I could have rebooted into recovery mode, dropped to root shell, and reset the root password on the boxes I manage. This is a 30-second procedure. Every hypervisor has some method to allow admins to boot guests into single user. ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: CppCheck Report on vm140 Not Running
Hi Luke, *, On Tue, Apr 6, 2021 at 3:42 PM Luke Benes wrote: > The script to email the dev list with the updated CppCheck Report at has > stopped running. back in June 2020… so as pointed out: not really actively used by people. > Could we get this restarted? Recreating it is always an option if there are volunteers maintaining it. But I certainly am not interested for getting blamed that service xy is failing for x weeks anymore like it has been the case with other services in the past. Infra will be more trigger happy to just kill unmaintained VMs. Also not interested in setting up specially handcrafted build environments. If it cannot be setup on a centos7 or debian 10 with standard packages, any setup required needs to be either added to lode or documented properly. But also don't expect Infra team to browse the list for random infra-requests. For larger stuff a redmine ticket is the best way, for smaller bits here and there #tdf-infra on IRC or mail to hostmas...@documentfoundation.org. So if Maarten is willing to take patronage of that, I'm happy to setup a hetzner cloud VM with the jenkins-buildslave setup as a playground. But let me restate: If manual setup steps are required, they need to be properly documented. It must be possible to set up the machine with salt states (distro packages, adding files, symlinks,...) or must be contained as a recipe in lode (if tools are required in a newer version than is available in centos7/epel/devtools repos) But given that the job had stopped so many months ago without anyone complaining - not sure how useful it really would be. ciao Christian -- Christian Lohmaier, Release Engineer Tel: +49 30 5557992-60 | IRC: cloph on Freenode The Document Foundation, Kurfürstendamm 188, 10707 Berlin, DE Gemeinnützige rechtsfähige Stiftung des bürgerlichen Rechts Legal details: http://www.documentfoundation.org/imprint ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: CppCheck Report on vm140 Not Running
Hi Luke, On Sat, Apr 24, 2021 at 03:55:50PM +, Luke Benes wrote: > How common is this? A check of > > git rev-list --all --grep='cppcheck' -i --count > 1847 > > Grepping shows a good mix of usual suspects and many GSoC students > looking for easy hacks. More importantly there were almost none since > the CppCheck script stopped working. I just described my workflow. :-) In my case, if I run these linters on a chunk of code I just wrote, I run them before pushing my commit, so typically it's not mentioned in the commit message. Anyhow, I did not argue that such a service is useless, you just asked why nobody complained when the service stopped, and I just mentioned one possible reason. Regards, Miklos ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: CppCheck Report on vm140 Not Running
> Yes, at least that's how I use these linters. How common is this? A check of > git rev-list --all --grep='cppcheck' -i --count 1847 Grepping shows a good mix of usual suspects and many GSoC students looking for easy hacks. More importantly there were almost none since the CppCheck script stopped working. ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: CppCheck Report on vm140 Not Running
Hi Maarten, On Thu, Apr 22, 2021 at 06:30:57PM +0200, Maarten Hoes wrote: > That may be true when running cppcheck on the entire codebase, but I now > assume (please correct me if I'm wrong here) that this is not how people > use the tool. Instead, cppcheck is likely to be run on a single file which > has just been modified, which takes second(s) at most. Yes, at least that's how I use these linters. There may be other valid workflows, e.g. some easy hacker actively looking for issues in files nobody touched recently. Regards, Miklos ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: CppCheck Report on vm140 Not Running
Now that I think of it, On Thu, Apr 22, 2021 at 2:51 PM Luke Benes wrote: > > I'm more interested in running such tools locally > > Depending on the configuration, cppcheck scans can take 8+ hours to run. > > That may be true when running cppcheck on the entire codebase, but I now assume (please correct me if I'm wrong here) that this is not how people use the tool. Instead, cppcheck is likely to be run on a single file which has just been modified, which takes second(s) at most. - Maarten ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: CppCheck Report on vm140 Not Running
Hi, On Thu, Apr 22, 2021 at 9:14 AM Miklos Vajna wrote: > I'm more interested in running such tools locally > > < - snip -> > > The other problem with cppcheck is that it doesn't build on an existing > c++ parser from a compiler (not based on e.g. gcc or clang), so its > signal/noise ratio is lower than e.g. coverity or clang-tidy. > > Miklos, thank you for taking the time to explain things. I already concluded that there was little interest in the job, but now I also know why that is. (Also, although Luke had contacted me before his post to the mailing list, I didn't know up front that Luke was going to do so and in what manner. I may or may not agree with what he has stated.) On Thu, Apr 22, 2021 at 2:51 PM Luke Benes wrote: > > > The other problem with cppcheck is that it doesn't build on an existing > > c++ parser from a compiler > > https://www.kickstarter.com/projects/3300446/cppcheck-clang-import > > They are working on a clang AST importer. Experimental support has already > landed. > > Which just got removed again: https://github.com/danmar/cppcheck/commit/207361b174102f38909aec9a996b944ecb370464 - Maarten ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: CppCheck Report on vm140 Not Running
> I'm more interested in running such tools locally Depending on the configuration, cppcheck scans can take 8+ hours to run. This can be greatly reduce with parallel checks at the cost of several checks being disabled. > The other problem with cppcheck is that it doesn't build on an existing c++ parser from a compiler https://www.kickstarter.com/projects/3300446/cppcheck-clang-import They are working on a clang AST importer. Experimental support has already landed. > its signal/noise ratio is lower than e.g. coverity or clang-tidy. A couple of years ago, I investigated this issue and found that cppcheck is missing include locations. When I provided those, the number of errors reported dropped from 9000+ to several hundred. https://wiki.documentfoundation.org/Development/Cppcheck At the time I manually generated include paths. Another option would be to use the ide-integration scripts to generate these. -Luke ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: CppCheck Report on vm140 Not Running
Hi Luke, On Wed, Apr 21, 2021 at 02:16:21PM +, Luke Benes wrote: > I spoke to the author of the cppcheck script, Maarten Hoes, about this > issue. We would like to know why no one has responded to our emails > about the cppcheck service being down. Are devs not interested in > cppcheck's results, is it the high false positives, or is it that no > one knows how to help get it back online? I'm more interested in running such tools locally, so whenever I fix an error/warning from such a tool, I can do have these steps: 1) run the tool -> see the result 2) attempt to fix it 3) run the tool again -> see the result disappearing If the tool is not running locally, 2) will be a blind fix, which is not ideal. E.g. IWYU has a wrapper in core.git, so you can run: bin/find-unneeded-includes path/to/test.cxx and it'll show you the IWYU result for that .cxx file. It is possible that such a wrapper for cppcheck would be also useful, and would require less maintenance that a full service. The other problem with cppcheck is that it doesn't build on an existing c++ parser from a compiler (not based on e.g. gcc or clang), so its signal/noise ratio is lower than e.g. coverity or clang-tidy. This is of course not your fault, but it explains if the interest is lower than what you expect. Regards, Miklos ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: CppCheck Report on vm140 Not Running
I spoke to the author of the cppcheck script, Maarten Hoes, about this issue. We would like to know why no one has responded to our emails about the cppcheck service being down. Are devs not interested in cppcheck's results, is it the high false positives, or is it that no one knows how to help get it back online? Maarten has lost his ssh-key to vm140 and needs support from the tdf infrastructure team to get back in. IF there is sufficient developer interest,he has generously offered to setup a new vm for the script. I too would be willing, if I could get a tdf infra team commitment to provide minimal support. (Hours of volunteered time were wasted on a ESC approved 32-bit linux tinderbox, because I could not get my quested answered on registering a lode tinderbox after I got the manual builds working) ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: CppCheck Report on vm140 Not Running
I dug a little deeper and the last error report has an "unexpected end of file" when I tried to decompress it.[1] Is vm140 out of free space? -Luke [1] https://dev-builds.libreoffice.org/cppcheck_reports/cppcheck_reports.tar.bz2 ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice
CppCheck Report on vm140 Not Running
The script to email the dev list with the updated CppCheck Report at has stopped running. Could we get this restarted? While it was not perfect, a quick grep of the commit log shows hundred of fixed bugs that it helped highlight. The script was running on host vm140 as /home/buildslave/source/dev-tools/cppcheck/cppcheck-report.sh And found at https://gerrit.libreoffice.org/plugins/gitiles/dev-tools/+/master/cppcheck/cppcheck-report.sh Thanks, -Luke ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/libreoffice