Re: Iridium vs Chromium

[Elias M. Mariani]

Actually, I was just checking the ports-changes mailing-list, and the
sync between Iridium and Chromium made me ask this.
In any case, OpenBSD ports has nothing to do with this question. I ask
here just because the OpenBSD community has a better view of this
things. And (so far) they had made interesting points when asking
about other things (Security Cameras for example).

> like the lagg is many months.
> are wrong about there being a lagg.  By telling the world the chromium
Is "lagg" different from "lag" or just a typo? Honest question.


On Sun, Apr 12, 2020 at 1:11 PM Theo de Raadt  wrote:
>
> Raymond, David  wrote:
>
> > That said, I am a bit nervous about OpenBSD's lags in
> > keeping up with browser security fixes.
>
> It isn't that simple.
>
> They don't ship security fixes standalone.  Instead, they ship a mix of
> new changes *and* fixes.  Lots of new unrelated changes, and only a few
> security fixes.  These fixes cannot be plausibly seperated out, as doing
> such a seperate procedure would increase the development workload, and
> increase the update lag.  So instead this software is accepted from mainstream
> on the assumption of their best effort, and then the following happens:
>
> The large changesets requires evaluation and verification, to ensure it
> still works with the pledge/unveil changes.  The pledge/unveil changes
> introduce a tighter sandbox than other operating systems have.  Quite
> often, upstream performs operations in the wrong order, and robert finds
> this out during test.  This friction slows the development a little.
>
> But I believe you are completely wrong about how long this lag is.
> You are being inaccurate because you don't know, and making it sound
> like the lagg is many months.
>
> 81.0.4044.92 came out 5 days ago, and here you can see it enter the ports
> tree 2 days ago.
>
> 1 hour ago it was replaced with a newer update.
>
> date: 2020/04/10 18:51:30;  author: robert
> update to 81.0.4044.92;
> date: 2020/04/03 13:44:40;  author: robert
> update to 80.0.3987.163
> date: 2020/04/01 12:32:05;  author: robert
> update to chromium-80.0.3987.162;
> date: 2020/03/21 14:08:01;  author: robert
> update to 80.0.3987.149 and apply the following changes:
> date: 2020/03/11 23:57:03;  author: espie
> date: 2020/03/04 15:44:17;  author: robert
> update to 80.0.3987.132 and fix the component flavor while here
> date: 2020/02/22 12:33:20;  author: robert
> update to 80.0.3987.116;
> date: 2020/01/17 20:43:38;  author: robert
> update to 79.0.3945.130
> date: 2020/01/08 14:43:32;  author: robert
> update to 79.0.3945.117
> date: 2019/12/18 09:01:35;  author: robert
> update to 79.0.3945.88
> date: 2019/12/15 12:03:46;  author: robert
> update to 79.0.3945.79
> date: 2019/11/20 18:26:30;  author: robert
> update to 78.0.3904.106
> date: 2019/11/07 10:47:41;  author: robert
> update to 78.0.3904.97
> date: 2019/11/05 22:30:26;  author: robert
> update to 78.0.3904.87
> date: 2019/10/22 18:35:43;  author: robert
> update to 77.0.3865.120 and make sure to use HW_NCPUONLINE instead of HW_NCPU
>
> You can pick through that list and compare the dates to the
> pledge/unveil adaptations commited into the tree.  It appears to move
> very rapidly, more rapidly than the average port.  The changes don't
> neccessarily make it into -stable and -stable packages, but *we never
> promised that*, and this specific pledge/unveil-using application is now
> using API that didn't exist in 6.6.
>
> > (I'm not criticizing -- I understand that ...
>
> Yes, you are are criticizing.  And with inaccurate statements.  And you
> are wrong about there being a lagg.  By telling the world the chromium
> openbsd effort is "slow", you are being an innaccurate downer.

Iridium vs Chromium

[Elias M. Mariani]

I'm not much of a browser savy guy.
Is Iridium really safer than Chromium?
Leaving aside the "Google is tracking you!".

Any recommendations on the browser front on performance, security and
compatibility?
I've been using Chrome and Chromium for years, but maybe there are
better alternatives that I'm unaware of...

Cheers.
Elias.

Re: Advice on Security Cameras

[Elias M. Mariani]

Thanks all for the help.

I will check out Zoneminder and the cameras that you have recommended.

> What do you want to do from the Android / browser?
Just look at the cameras from outside the house, I don't need any type
of functionality besides that.

Advice on Security Cameras

[Elias M. Mariani]

Hi list,
I'm thinking in installing some cameras in my private home, I have
been looking for solutions, my concern is that I wish to be able to
look the videos from outside the house and I'm a little paranoid about
the quality of the software that the different vendors use. I have
seen clusters of camaras that only work over ActiveX...
I know that is a little off-topic but maybe someone knows about a good
brand of cameras.
Of-course one can always set a VPN tunnel trough OpenBSD for the
security matter, OpenVPN works on Android so is easy to access from a
smartphone. But I would prefer to have a single secure service running
that adding a layer of complexity with the VPN.

I'm looking for:
- Not overpriced cameras.
- They don't need to be "external cameras", they will be covered under a roof.
- I need to set at least 4, so I need them to be accessible from a
single platform.
- Android / Browser friendly (not only IE plz...)
- WiFi is not needed, I have a 12v supply and Ethernet connections for
each camera.
- Good video quality but I'm not looking for anything super great...
- the ability to centralize recording and access to view the cameras is a must.

Again, sorry for the off-topic but were would I find a better place to
ask about surveillance and security ? :D

Cheers and happy new year.
Elias.

Re: php 5.6 and php 7 using php-fpm ?

[Elias M. Mariani]

That hit the spot.
Thanks!

2018-08-16 18:57 GMT-03:00 Stuart Henderson :
> To run the two versions concurrently, use the rc.conf.local flags variables
> (php56_fpm_flags/php70_fpm_flags) to give them different config files (-y
> /path/to/fpm.conf).
>
> --
> Sent from a phone, apologies for poor formatting.
>
>
> On 16 August 2018 22:09:57 "Elias M. Mariani" 
> wrote:
>
>> Hi,
>> Somebody knows how to set up 2 different socks, one with php56 and
>> another with php70 ?
>> Yo can just run
>> rcctl start php56_fpm php70_fpm
>> Because they would use the same fpm.sock.
>> And this is configured in /etc/php-fpm.conf, I did not found another
>> place to configure this.
>>
>> Cheers.
>> Elias.
>
>
>
>

php 5.6 and php 7 using php-fpm ?

[Elias M. Mariani]

Hi,
Somebody knows how to set up 2 different socks, one with php56 and
another with php70 ?
Yo can just run
rcctl start php56_fpm php70_fpm
Because they would use the same fpm.sock.
And this is configured in /etc/php-fpm.conf, I did not found another
place to configure this.

Cheers.
Elias.

NSD update

[Elias M. Mariani]

I saw a week ago this commit from florian:
https://marc.info/?l=openbsd-cvs=153294104203261

Shouldn't this apply also to 6.3 ?

Just a doubt not a complain.
Elias.

Re: httpd rewrite, what I'm missing?

[Elias M. Mariani]

Just a breaktrough, but the problem persists:

-
#example of directory that doesn't need to use rewrite.
location match "/themes/(.*)/assets/(.*)" {
request no rewrite
}

#index.php is the entry point of the application and runs the php fastcgi.
location match "index.php" {
fastcgi socket "/run/php-fpm.sock"
}

#all others locations rewrite to /index.php
location match "(.*)" {
request rewrite "/index.php$REQUEST_URI"
}
-

The problem is that when rewriting to index.php in this manner the
links are formatted in the same way as the rewrite, is there a way
apart from "request strip" that I can use to change the $REQUEST_URI
so the fastcgi socket gets the value with the "/index.php" striped ?

cheers.
Elias.

2018-07-17 17:13 GMT-03:00 Elias M. Mariani :
> Hi,
> I'm trying to adapt the rewrite rules of OctoberCMS to httpd.conf to
> avoid using apache-httpd.
>
> Now everything is working ok with this rules:
>
> -
> #example of directory that doesn't need to use rewrite.
> location match "/themes/(.*)/assets/(.*)" {
> request no rewrite
> }
>
> #index.php is the entry point of the application and runs the php fastcgi.
> location match "index.php" {
> fastcgi socket "/run/php-fpm.sock"
> }
>
> #all others locations rewrite to /index.php
> location match "(.*)" {
> request rewrite "/index.php"
> }
> -
>
> But for some reason the URLs are working like this:
> example.com/forums/openbsd
> Works ok, rewrites and everything is OK, but for some reason all the
> links point to:
> example.com/index.php/forums/openbsd
> That also works the same as is the first case.
>
> Now, this URLs are defined by the application, I guess that the
> application understands that the base_url (I made the name up...) is
> example.com/index.php and not example.com ?
>
> Might this be an incompatibility between httpd and apache-httpd?
>
> Cheers.
> Elias.

httpd rewrite, what I'm missing?

[Elias M. Mariani]

Hi,
I'm trying to adapt the rewrite rules of OctoberCMS to httpd.conf to
avoid using apache-httpd.

Now everything is working ok with this rules:

-
#example of directory that doesn't need to use rewrite.
location match "/themes/(.*)/assets/(.*)" {
request no rewrite
}

#index.php is the entry point of the application and runs the php fastcgi.
location match "index.php" {
fastcgi socket "/run/php-fpm.sock"
}

#all others locations rewrite to /index.php
location match "(.*)" {
request rewrite "/index.php"
}
-

But for some reason the URLs are working like this:
example.com/forums/openbsd
Works ok, rewrites and everything is OK, but for some reason all the
links point to:
example.com/index.php/forums/openbsd
That also works the same as is the first case.

Now, this URLs are defined by the application, I guess that the
application understands that the base_url (I made the name up...) is
example.com/index.php and not example.com ?

Might this be an incompatibility between httpd and apache-httpd?

Cheers.
Elias.

Re: Mail Server

[Elias M. Mariani]

Well, the thing is that the accounts are managed by our clients, so I
need a platform where they can easy administrate the accounts of their
domains and they are common people, it has to be easy.
I guess that using Postfix would solve the SMTP part.

2018-07-06 20:51 GMT-03:00  :
>
> On Jul 6, 2018 5:56 PM, "Elias M. Mariani"  wrote:
>>
>> Hello,
>> I'm moving some sites that my friends and I handle from a cheap
>> webserver to a cloud server with OpenBSD.
>> I have the webserver part figured out (httpd+mariadb+php7).
>> But I need to choose how to handle the mails.
>> I'm mostly worried that the mail accounts must be managed by some of
>> our users through the web.
>> So I need to use something like postfixadmin or ViMbAdmin, that is, a
>> web application to handle the accounts.
>> Also I dont have much knowledge about what the market has to offer...
>> Should I use OpenSMTPD or I need some big MTA like postfix ?
>> I also need IMAP so can I use OpenSMTPD+dovecot+ViMbAdmin?
>>
>> This is a kind of open question, maybe someone savy in mail servers
>> can guide me in what to choose.
>> The amount of mails in and out is very small, so, I'm not looking for
>> performance, just for security, reliability and, truth be told...
>> practicality, to many components to handle a mail... It should be
>> simpler... ;(
>>
>> Cheers.
>> Elias.
>>
>
> Are you going to need to add and remove accounts on a daily basis? I believe 
> vimbadmin only works with postfix. However, if you are just setting up a few 
> accounts that won't need to change immediately semi frequently then opensmtpd 
> will likely be much easier for you to set up.

Mail Server

[Elias M. Mariani]

Hello,
I'm moving some sites that my friends and I handle from a cheap
webserver to a cloud server with OpenBSD.
I have the webserver part figured out (httpd+mariadb+php7).
But I need to choose how to handle the mails.
I'm mostly worried that the mail accounts must be managed by some of
our users through the web.
So I need to use something like postfixadmin or ViMbAdmin, that is, a
web application to handle the accounts.
Also I dont have much knowledge about what the market has to offer...
Should I use OpenSMTPD or I need some big MTA like postfix ?
I also need IMAP so can I use OpenSMTPD+dovecot+ViMbAdmin?

This is a kind of open question, maybe someone savy in mail servers
can guide me in what to choose.
The amount of mails in and out is very small, so, I'm not looking for
performance, just for security, reliability and, truth be told...
practicality, to many components to handle a mail... It should be
simpler... ;(

Cheers.
Elias.

Re: httpd chroot outbound

[Elias M. Mariani]

Thanks for the help guys,
I was linking, not copying resolv.conf

Fixed.
Thanks again.

Elias.

2018-06-25 13:59 GMT-03:00 Scott Vanderbilt :
> On 6/25/2018 9:37 AM, Elias M. Mariani wrote:
>
>> Does anybody knows what is needed to allow php to retrieve files while
>> under httpd chrooted ?
>> I recall the need of /etc/resolv.conf on the jail but that didn't work.
>
>
> Also: http://php.net/manual/en/install.unix.openbsd.php

httpd chroot outbound

[Elias M. Mariani]

Hi.
Does anybody knows what is needed to allow php to retrieve files while
under httpd chrooted ?
I recall the need of /etc/resolv.conf on the jail but that didn't work.

Cheers.
Elias.

Re: OpenBSD performance upgrade

[Elias M. Mariani]

It uses AVX2, so... thanks for the extra seconds. :D

Cheers.
Elias.

2018-06-09 0:43 GMT-03:00 Philip Guenther :
> On Fri, Jun 8, 2018 at 10:13 AM Elias M. Mariani 
> wrote:
>>
>> I usually run long computations on OpenBSD-current, in the last few
>> days I see an upgrade in the performance of the process (in this case
>> I have 6 threads running a very optimized assembler code).
>> Each iteration of the code was about 14 sec. and now is around 13 sec.
>> Don't mind the computation, the question is more about if something
>> was changed (and if so what) to hit the performance so much, the
>> compilation is the same as before, so the code did not change and the
>> software does not use anything from ports, only the standard C
>> library.
>>
>> A more accurate description would be that the threads are more
>> homogeneous in relation with the iterations time, maybe one thread
>> suddenly give a 11 sec. result, and another 15, probably related with
>> thread reallocations, now I have a steady time in all the threads and
>> in average the numbers are better.
>> Anyways... Better is better. Just curious to know why is working better.
>> :D
>
>
> If the "optimized assembler"  that the threads are running uses AVX or
> similar "extended CPU state" extensions then the improvement is almost
> certainly from my switch amd64 from "lazy FPU switching" to what I'll call
> "semi-eager switching", where the current thread's registers are always
> ensured to be loaded before returning to userspace, eliminating the need for
> extra userspace->kernel->userspace transitions and IPIs to load the
> registers in the current CPU.
>
> Glad to hear it's such a large improvement for your processing.  Enjoy, and
> remember to tip your OS vendor!  
>
>
> Philip Guenther
>

Re: OpenBSD performance upgrade

[Elias M. Mariani]

Forgot to add: amd64, last version of current.

2018-06-08 14:12 GMT-03:00 Elias M. Mariani :
> Hi,
> Just another silly question.
> I usually run long computations on OpenBSD-current, in the last few
> days I see an upgrade in the performance of the process (in this case
> I have 6 threads running a very optimized assembler code).
> Each iteration of the code was about 14 sec. and now is around 13 sec.
> Don't mind the computation, the question is more about if something
> was changed (and if so what) to hit the performance so much, the
> compilation is the same as before, so the code did not change and the
> software does not use anything from ports, only the standard C
> library.
> A more accurate description would be that the threads are more
> homogeneous in relation with the iterations time, maybe one thread
> suddenly give a 11 sec. result, and another 15, probably related with
> thread reallocations, now I have a steady time in all the threads and
> in average the numbers are better.
> Anyways... Better is better. Just curious to know why is working better. :D
>
> Cheers.
> Elias,

OpenBSD performance upgrade

[Elias M. Mariani]

Hi,
Just another silly question.
I usually run long computations on OpenBSD-current, in the last few
days I see an upgrade in the performance of the process (in this case
I have 6 threads running a very optimized assembler code).
Each iteration of the code was about 14 sec. and now is around 13 sec.
Don't mind the computation, the question is more about if something
was changed (and if so what) to hit the performance so much, the
compilation is the same as before, so the code did not change and the
software does not use anything from ports, only the standard C
library.
A more accurate description would be that the threads are more
homogeneous in relation with the iterations time, maybe one thread
suddenly give a 11 sec. result, and another 15, probably related with
thread reallocations, now I have a steady time in all the threads and
in average the numbers are better.
Anyways... Better is better. Just curious to know why is working better. :D

Cheers.
Elias,

Re: build and ports mismatching ?

[Elias M. Mariani]

Great explanation,
Thanks.
Elias.

2018-05-24 15:59 GMT-03:00 Sebastian Benoit <benoit-li...@fb12.de>:
> Elias M. Mariani(marianiel...@gmail.com) on 2018.05.24 15:45:15 -0300:
>> Thanks Sebastian for the reply,
>> I do follow source-changes, just that I don't understand if it's
>> common for a library to have 2 version numbers.
>
> There is the old version number(s) and the newest one.
> If you check your /usr/lib and /usr/local/lib you will find
> multiple versions of the same library.
>
> (The old ones are no longer needed when all software that was build agains
> them has been updated.)
>
>> Or if the ports/snapshots system is made with some inner logic to
>> avoid this kind of conflict.
>
> When you build software, its linked agains the newest version of the lib
> thats available on your system.
>
> Base system snapshots (including xenocara) and packages are not build
> at the same time. Package builds use the latest snapshot, but since
> building packages takes a day or so (depending on the arcitecture and
> build machines available) and distribution of the packages to the mirrors
> also tkes time, there is a time lag between the newer version of a base
> library available and the package that uses it being available.
>
> Add to that that you (as user) probably do not catch every library bump of
> base system libries when updating your machine, you can get into the
> situation that base has a newer lib than what a package needs, but the older
> lib that the package needs is newer than the old version thats still on
> your system because you skipped that.
>
> The "problem" (it's not if you know what you are doing) could be solved by
> making sure package builds and base are always in sync. But we do not want
> to do that because it slows down development.
>
>> I mean, the ports did not install the library, that means that one of
>> the packages of x did it.
>> Cheers.
>> Elias.
>>
>> 2018-05-24 15:26 GMT-03:00 Sebastian Benoit <benoit-li...@fb12.de>:
>> > Elias M. Mariani(marianiel...@gmail.com) on 2018.05.24 14:22:35 -0300:
>> >> Hi,
>> >> I noticed just now a couple of errors after updating from
>> >> snapshots/amd64 (22/05) and updating the packages with pkg_add -u
>> >> (24/05) indicating a mismatch in some library, I think it was
>> >> libfreetype.so.28.2 vs 29.0 or something like that.
>> >> I have both in /usr/X11R6/lib/.
>> >> I'm not familiar with the correlation between snapshots and ports. And
>> >> less about C libraries...
>> >
>> > matthieu@ updated xeoncara stuff on the 22nd.
>> >
>> > Either your snapshot does not have the updates yet, but the packages you 
>> > try
>> > to install have been build with the new library, or you have the new X
>> > libraries, but the snaps lag a bit behind. Probably the later.
>> >
>> > Wait for a day and try pkg_add again, and/or update to a newer snapshot 
>> > too.
>> >
>> > If you run current, you should follow source-changes
>> > (http://www.openbsd.org/mail.html), otherwise you might run into problems
>> > like this. Of course it also requires understanding the commit and its
>> > possible impact.
>>
>
> --

Re: build and ports mismatching ?

[Elias M. Mariani]

Thanks Sebastian for the reply,
I do follow source-changes, just that I don't understand if it's
common for a library to have 2 version numbers.
Or if the ports/snapshots system is made with some inner logic to
avoid this kind of conflict.
I mean, the ports did not install the library, that means that one of
the packages of x did it.
Cheers.
Elias.

2018-05-24 15:26 GMT-03:00 Sebastian Benoit <benoit-li...@fb12.de>:
> Elias M. Mariani(marianiel...@gmail.com) on 2018.05.24 14:22:35 -0300:
>> Hi,
>> I noticed just now a couple of errors after updating from
>> snapshots/amd64 (22/05) and updating the packages with pkg_add -u
>> (24/05) indicating a mismatch in some library, I think it was
>> libfreetype.so.28.2 vs 29.0 or something like that.
>> I have both in /usr/X11R6/lib/.
>> I'm not familiar with the correlation between snapshots and ports. And
>> less about C libraries...
>
> matthieu@ updated xeoncara stuff on the 22nd.
>
> Either your snapshot does not have the updates yet, but the packages you try
> to install have been build with the new library, or you have the new X
> libraries, but the snaps lag a bit behind. Probably the later.
>
> Wait for a day and try pkg_add again, and/or update to a newer snapshot too.
>
> If you run current, you should follow source-changes
> (http://www.openbsd.org/mail.html), otherwise you might run into problems
> like this. Of course it also requires understanding the commit and its
> possible impact.

build and ports mismatching ?

[Elias M. Mariani]

Hi,
I noticed just now a couple of errors after updating from
snapshots/amd64 (22/05) and updating the packages with pkg_add -u
(24/05) indicating a mismatch in some library, I think it was
libfreetype.so.28.2 vs 29.0 or something like that.
I have both in /usr/X11R6/lib/.
I'm not familiar with the correlation between snapshots and ports. And
less about C libraries...
Though of giving a heads up just for the error that updating
ImageMagick give, ignore this message if this is common.

Cheers.
Elias.

Re: Building OpenBSD and ports VS installing from packages

[Elias M. Mariani]

Okey, thanks both for the help!
Elias.

Re: Building OpenBSD and ports VS installing from packages

[Elias M. Mariani]

Hi,
I understand that about the builds and packages.
I will re write my question in another form:
If I build, say, firefox on a i386 machine I get a package, and
another if I build firefox on amd64, they differ.
If I build firefox on an amd64 machine WITHOUT AVX support I get a
package, if now I build firefox in an amd64 machine WITH AVX support,
do I still get the same package ?
(firefox is a random pick)

Cheers.
Elias.

2018-05-21 18:24 GMT-03:00 IL Ka :
> Hello.
>
> OpenBSD team does not recommend to build anything that exists in packages.
>
>>>If so, building from ports would produce a different code?
> In most cases ports are not aware of your microarchitecture.
> See my question and Theo's answer.
>
> https://www.mail-archive.com/misc@openbsd.org/msg160878.html
>
>
>

Building OpenBSD and ports VS installing from packages

[Elias M. Mariani]

Hi,
I have this question in my mind for a time now, if I download OpenBSD
and install all the applications from packages do OpenBSD and the apps
use for example AVX512 ? I mean, if I understand correctly, the
compiler should optimize the code for a given set of instructions,
given that, for example, the amd64 distribution works on machines with
and without AVX I assume that the packages are built without AVX
support, is this correct?
If so, building from ports would produce a different code?
Is the same for the OpenBSD build? Would I get a better performance if
I compile from source instead of installing from the packages?

I know that is a silly question, is just that I'm not used to think in
terms of compiled languages, and I want to get this question out of my
head. xD

Elias.

Re: re0: watchdog timeout on recent current

[Elias M. Mariani]

Had the same issue with a TP-LINK  TG-3269
https://www.tp-link.com/en/products/details/cat-11_TG-3269.html
On high loads it started to show the watchdog timeout.
It just used another adapter but there is clearly a bug with this one.
Elias.


2018-05-01 6:08 GMT-03:00 Anthony J. Bentley :
> Gregory Edigarov writes:
>> Hello everybody,
>>
>> ok, so here is the symptoms. the thing happens usually during the high
>> traffic, like when I am trying to watch video on a tv, which is
>> connected to my home server/router on re0 (it is the local interface).
>>
>> the video freezes immediately. something like ifconfig re0 down && pfctl
>> -Fst && ifconfig re0 up, hepls a bit but not every time, sometimes I
>> need to reboot.
>>
>> during the March and until the middle of April it was working fine, do I
>> think it was broklen quite recently.
>
> I've had this problem on an RTL8168G since I first got the machine three
> years ago. It has gotten much better; back then it happened a few times
> a day and locked up the machine solid. I still see it from time to time
> (including a couple of times this week) but it barely slows things down
> and ifconfig down/up is usually enough to bring it back.
> https://marc.info/?l=openbsd-misc=143701763323995=2
>
> I have no idea where to even begin to make an effective bug report.
> Unfortunately in all this time I've never figured out how to reproduce
> it either--just that it happens occasionally under heavy network load.
>

Re: swi-pl : tiny edition!

[Elias M. Mariani]

Hi,
I think that you can build your own package of swi-prolog by modifying
the Makefile in the ports tree, just read here:
http://www.swi-prolog.org/build/prerequisites.html
About what each of the dependencies do, remove the ones that you don't
want from LIB_DEPENDS and make a new PLIST for the package.
If not, you should look for the maintainer and ask him if it can make
pseudo-flavors for optional dependencies.
Just out of curiosity, why do you need a smaller version ?
Elias.

2018-04-30 3:38 GMT-03:00 Mayuresh Kathe :
> i don't have the skills nor the experience to accomplish this, so here.
> can there be a swi-pl-tiny edition of swi-prolog?
> as it stands today, swi-prolog has a whole lot of dependencies, as here;
> gmp, libexecinfo, pcre, ossp-uuid, jpeg, bzip2, lz, xz, libarchive, db,
> iodbc.
> it would be really helpful to have that tiny edition for a project i am
> working towards for and under openbsd.
> thanks.
>

Re: CPU Affinity

[Elias M. Mariani]

Thanks for the reply Peter.
May I ask why is not available to userland?
Is just a developers decision or is because no one is interested in
that functionality?

Elias.

2018-04-30 3:32 GMT-03:00 Peter Hessler <phess...@theapt.org>:
> On 2018 Apr 29 (Sun) at 22:07:18 -0300 (-0300), Elias M. Mariani wrote:
> :Hi,
> :I was trying to port mprime to OpenBSD.
> :The main issue is not finding any way to set affinity on cores.
> :Searching for how to do this on OpenBSD bring this result in undeadly:
> :http://undeadly.org/cgi?action=article=20090324210236
> :
> :Is CPU Affinity dropped out of OpenBSD for some reason?
> :
> :Elias.
> :
>
> CPU Affinity is available inside the kernel only, and is not exposed to
> userland.
>
> There are no plans to make it available to userland.
>
>
> --
> Pardon this fortune.  Database under reconstruction.
>

Re: CPU Affinity

[Elias M. Mariani]

Hi Anton,
The why is:
An increase in performance.
In this case I want to use some cores to do a very large operation and
it will be faster if the same core continues the same operation.
Maybe there's a library for doing so, or a way that I dont know in OpenBSD.
Now, this cases come not very often, and is debatible (probably
already debated in this proyect) that if that's the case you should
run a variation of the scheduler or use other OS.
And is true...
Now, in this moment I see a lot of activity around vmd/vmm, I have
been using it myself, is great, I love it, but my guess is that you
can get more performance on the guests OS if you set the affinity of
the process.
Maybe this is already done, or may will be... If that's the case
either you added functionality to the OS for supporting affinity or
you made vmm part of the OS, and there is the back question.
If you add vmm as part of the OS you are "extending" the system duty
to be a host, if you add affinity support you are doing the same...
Anyways, the proyect clearly knows itself and if there is no affinity
support is OK. Still loving OpenBSD.
The low punch would be discuss about "nice" support, but lets leave
that aside, I dont want a fight like in other post about philosophy
and low punches.

Now in MY PARTICULAR CASE, I usually do large math calculations, I've
been using other OSes but I love OpenBSD, so I ported my most used
tool, spyder (for writing python).
I love to test algorithms and things for a large time and things like
that. That is my taste, and is not an obligation for OpenBSD to
support affinity just because I want (Its obvious), but I see lots of
applications for "user level programs", for example vmm, pf, and
others components, I dont know if PF and VMM are "user level
programs", if they are you are already doing things that might be seen
as "not OS duty", if they are not, they would benefit probably of
affinity.

Just my opinion, again, dont do of this remarks something that it is not...
I just kindly ask about a funcionality, because I want to use my
favorite OS, made right, and test algorithms and look for big prime
numbers on it. If there is no funcionality, then, Its my problem...
Elias.

2018-04-29 23:15 GMT-03:00  <li...@wrant.com>:
> Sun, 29 Apr 2018 22:07:18 -0300 "Elias M. Mariani"
> <marianiel...@gmail.com>
>> Hi,
>> I was trying to port mprime to OpenBSD.
>> The main issue is not finding any way to set affinity on cores.
>> Searching for how to do this on OpenBSD bring this result in undeadly:
>> http://undeadly.org/cgi?action=article=20090324210236
>>
>> Is CPU Affinity dropped out of OpenBSD for some reason?
>>
>> Elias.
>>
>
> Hi Elias,
>
> Why do you want to do the operating system duty in user level programs?
>
> Kind regards,
> Anton Lazarov

CPU Affinity

[Elias M. Mariani]

Hi,
I was trying to port mprime to OpenBSD.
The main issue is not finding any way to set affinity on cores.
Searching for how to do this on OpenBSD bring this result in undeadly:
http://undeadly.org/cgi?action=article=20090324210236

Is CPU Affinity dropped out of OpenBSD for some reason?

Elias.