Summary: | Compiling media-gfx/k3d-0.6.7.0-r1 with SSP-enabled gcc-3.4.6 results in immediate memory saturation, failure | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Gordon Malm (RETIRED) <gengor> |
Component: | Hardened | Assignee: | The Gentoo Linux Hardened Team <hardened> |
Status: | RESOLVED FIXED | ||
Severity: | normal | CC: | atoth, graphics+disabled, toolchain |
Priority: | High | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Bug Depends on: | |||
Bug Blocks: | 135265 |
Description
Gordon Malm (RETIRED)
2008-06-30 00:19:24 UTC
if this program is C++ based. Might I suggest to add -fno-fstack-protector-all to "CXXFLAGS" May I suggest to give https://hardened.gentooexperimental.org/secure/ a spin. Another tip: -fno-strict-overflow. Regards: Dw. Fixed in CVS. Wait an hour and emerge --sync. (In reply to comment #3) > Fixed in CVS. Wait an hour and emerge --sync. > I hope that the applied -fno-stack-protector "fix" will affect only hardened toolchains using pre-4.x gcc. Did you give a mf spin to the experimental toolchain before "resolving" the bug (as I suggested)? Regards: Dw. > I hope that the applied -fno-stack-protector "fix" will affect only hardened > toolchains using pre-4.x gcc. Yes, it only affects <gcc-4. Furthermore, I've added an append-cxxflags a few days ago, which is utilized here. So SSP is only turned off for the C++ portions when using gcc-3.x. You're a smart guy, try looking at the Changelog or the ebuild. And yes, the bug is "fixed" because the package compiles now. gcc-3.4.6 C++ SSP implementation won't be fixed and neither will K3D. The bugs like this are starting to pile up and its a rather nasty failure that can potentially fool the OOM killer into killing off other processes. As much as it sucks, better to just have the packages compile until a white knight comes along and fixes everything. > Did you give a mf spin to the experimental toolchain before "resolving" the bug > (as I suggested)? > Who do you think has been pulling the bits into mainline and fixing bugs in it? solar (mainly) and myself. Zorry deserves big credit++ for getting it to the point it is at and finding/fixing bugs in it as well. > Another tip: -fno-strict-overflow. Who do you think this "tip" came from? Obviously, my patience has worn out for your attitude and snarky "mf" comments. (In reply to comment #5) > Who do you think has been pulling the bits into mainline and fixing bugs in it? > solar (mainly) and myself. Zorry deserves big credit++ for getting it to the > point it is at and finding/fixing bugs in it as well. I should apologize in public. It's a shame what I did by posting that unintelligent reply. I try to make no excuses, because there's no good reason. I have to admit, I didn't intended to insult you. I made a big mistake. Yesterday was a bad day for me, because I had to face with my own hickness. I'd like to let you know: I appraise your efforts and all others working on hardened. I still would like to be a valuable member of the community. Maybe these words won't make you forget. But to prove my willingness to make you forgive me: please let me know if there is something I can help you with, like hunting down bugs or something. Apart from that I'll watch out not to send silly post in the early morning, while failing to take a look at on the sources in a bad mood. Truly sorry: Dwokfur Sorry for the bugspam, so others please bare with me for one more public exchange. Dwokfur, I am sorry I snapped at you. You have contributed in the past and I appreciate that. It's an excuse, but to keep it short - there's a lot of not-so-great things going on for me/my family personally right now, so sometimes my mood is not so good and patience gets a bit thin. You don't know how much I'd love gcc 3.4.6's stack-protector implementation to work properly with C++ code. Unfortunately I'm not qualified to make that happen and Dr. Etoh disappeared long before I got here. I'm actually tempted to see if I can strip out SSP on C++ in gcc-3.4.6 and just fix it that way instead of editing ebuilds. It never worked properly anyway and afaik gcc-4 doesn't do SSP on C++ either. Adding -fno to CXXFLAGS in the default make.conf is a no-go for various reasons - in case you were going to suggest that. ;) Anyway, if you want to talk further or about bugs to be fixed, etc. please e-mail me. Thanks. |