-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hello,
I've looked into that issue a bit futher, and as far as I can test within the confines of a VM I can say that 1) I can still access the TTYs and 2) the OOM killer removes the offending process as soon as both RAM and Swap is full. Since your Swap is rather large it takes a long time to fill up, during which time the system constantly has to swap in and out programs you might be using. Aside from the obvious issue here you've apparently made a conscious effort to switch the system browser, as Debian by default installs Iceweasel, not Epiphany. That means it's not a combination that the Debian QA team had high on their list of priorities. Why is that important? You lament that you switched to Debian from another OS because you wanted stability. Given the high number of different DEs, browsers, mail clients, chat programs, ... in any given distribution the QA team can only test certain combinations, and make those as stable as possible. Since the browser from the GNOME project isn't included with the GNOME default install I'd wager it was exchanged because of bugs like this. But the system stability didn't suffer. The OOM killer removes the offending browser process once the memory is used up, and the system doesn't die. Once that happens, everything else is back to normal. Even a bug in the GUI can't (usually) kill the whole system. If you really want to protect yourself from issues like that in the future, either switch to a better maintained browser (Iceweasel is the default for Debian for example) and/or set a hard memory limit through limits.conf or control groups. Regards, /peter Am 27.10.2015 um 10:05 schrieb Abou Al Montacir: > On Tue, 2015-10-27 at 09:38 +0100, Peter Ludikovsky wrote: >> Hello, >> >> This is not a bug with Debian in particular, but with Epiphany >> in general. I could reproduce a similar behaviour, but managed to >> kill the browser process before the system became completely >> unresponsive. >> >> I suggest you file a bug report with the Gnome team: >> https://bugzilla.gnome.org/enter_bug.cgi?product=epiphany >> >>> > Thank Peter for you answer and you test. > > I fully understand that there is a bug in epiphany browser and that > one shall be reported both against the package as I consider this > serious and against the upstream bugzilla. > > That is not the point I want to discuss, but rather shall I open a > bug against general as the system is supposed to be robust enough > against any buggy user program. This particular capability of > Unixes system that make it possible for multiple users to share the > same computer, but also for the same user to ensure no program is > jeopardizing the others. > > Isn't Linux kernel supposed to care about this kind of issues? In > the past I knew that if a program was trying to allocate too much > memory, more that what the system could, it used to get a > segmentation fault, or similar signals, if a program enters an > infinite while loop it get anyway preempted to let other thread > run... All the nice features that make one migrate from a popular > OS to a nice one! >>> Can you please advise if I need to fill a bug against general? >>> PS: Please copy me as I'm not subscribed. > > -- Cheers, Abou Al Montacir -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.22 (MingW32) iQIcBAEBAgAGBQJWL3xZAAoJEM+6Ng5pbtyZLsgQAJK4J8/J1QQWGpeuKLOLIMLU z/mIszPnIuX5Ouo3vUM7YAcSW07D5cIg/6tuIhu3MiruSAgFNGzeXPergb8m1Tkp nJDOrxrNbhRxBYTmvqZ5mM/JoiErzVmRAw/JHjrpVvOxiJMu0XmAVo3tIre4QUsB MP6hhvu/5yfomP+aad83bhfh75m+zzXLoT3oAw7C9Qh85WfBWHwoC/SCAm5HydKt OImaIAN7eakdsJvprUtZHyM4PNYAT7RUviAlxER6PbaJS3XVOruy46vcUYCuOtxf 449BI2ifwMpHgTzcvgXhYd6qhS51UdSESw+Buo8YBdnGntuDA7nM19Lif+wYI6Qc F37NnfTxKu43L8DpnlFXQOzZ6N5WweNSF96ylMFJn1LxQywPKClMip1dk8uojw0H lr4lDix3JATcenzVWlLzrRfFdHq2yjM47euZ5elZjQI0wAcGdIsGABnBifO7Y6gI AdguQHDojGeCNzYnZCTpksyIIanXfoc3ip7pKEq2ZXPuiIWVlGlAq8cetNT1mXk+ RQPDgTdKlbjqdNf+fwWO9E9vuwKEz8qf40yETztm1Z586FU33EsOZZ3xiY+HNu6N 1gILihmXduF/i2x99Xg0oUPv132H/NmG21lJ+4tWGlTtV+Kc6Z8XaZZ+/gd2poq+ lkykv2JKh2Ap20Vu18F8 =AjMN -----END PGP SIGNATURE-----
