Linuxu­lat­or news

Today I com­mit­ted the up­dated linux aio stuff to p4. It’s a mod­ule now, so it can be loaded as needed. I also up­dated the p4 diff in the wiki, so any­one can down­load and test it (please do it!).

Ma­jor fea­tures in the p4 diff are fu­texes and TLS for amd64. So if you own an amd64 sys­tem please down­load the patch and give it a try. I’m in­ter­ested in re­ports for linux-fire­fox, skype, real­play­er and acror­ead on amd64.

There’s one known bug with the mmap be­ha­vi­or on amd64 (even in cur­rent). I don’t know if this only af­fects the LTP tests or real world ap­plic­a­tions, but be­cause there’s much time between the mmap com­mits and the re­port, I think it’s only a prob­lem for the re­gres­sion tests and not a prob­lem which shows up in typ­ic­al us­age scen­ari­os.

Some fixes for the linuxolat­or after test­ing by users

After the call for pub­lic test­ing we got some re­ports about LORs and a pan­ic. This happened after an ex­tens­ive par­al­lel com­pil­ing ses­sion of linux stuff on a SMP sys­tem for sev­er­al hours. Ro­man re­did some of the lock­ing and the fixes are in the tree. In­ter­me­di­ate patches already showed prom­ising res­ults. We’re wait­ing for the test of the com­mit­ted solu­tion now.

Yes­ter­day we also got re­ports about seg­faults with linux Java 5 and 6. This is un­der in­vest­ig­a­tion (any de­bug­ging help would be ap­pre­ci­ated, Java is a large beast).

I also no­ticed that we don’t have the res­ults of a LTP run of a nat­ive Linux sys­tem in the wiki. It would be very nice to have this, as it would al­low us to see broken test cases (Ro­man thinks the openat tests are flawed) or at least we see where it doesn’t mat­ter much that we don’t PASS.

An­oth­er sound de­veloper with per­force ac­cess

Apart from Kon­stantin (envy24* driver au­thor) I also got a p4 ac­count for Yur­iy (emu10kx driver au­thor). Let’s cross fin­gers for a lot of col­lab­or­a­tion in /​/​depot/​pro­jects/​sound­sys­tem/​ now…