As my comment on J5's blog is still awaiting moderation, I'll post the followup here.
What I feel wrong is to drop a working solution for one which despite a lot of work still makes so many people have no sound, or bad quality sound, or a machine too slow too read dvd, or…
Enforcing people to use it when it works for 99% of people might be good as the 1% may be fixed then. But currently I feel that much more than 1% people get issues with pulseaudio, and there are not enough people to fix them… (And some can’t be fixed, it will always add overhead so not powerful machines, like netbooks, will not be able to do things they can do when it is disabled).
I was in favor of pulseaudio being enabled by default on distros, because there was an easy way to disable it for people who had troubles with it and a few people got new functionalities.
Despite all the mails by lennart and others, I still see no real benefit in using pulseaudio for me (I have a laptop with 2 internal speakers, most of the time my sound is mute except when I have only one source of sound: when I listen some music or watch a movie, my brain can’t handle both at the same time), and have cases when it’s better to disable it.
For a long time I have been hearing that the issues are not in pulseaudio but in alsa or the apps. I believe it, but actually I don’t care. alsa has a lot of bugs but works fine, showing alsa bugs that will need years to get fixed, meaning that people won’t have sound during the next few years doesn’t make sense to me.
The desktop is supposed to be used, I don’t see the point of telling people “we broke it and you should keep an old version of the desktop or not use our desktop during the next few years until we stabilize this if you want a decent experience”
Mails like this does not help too.
« Before we can properly fix this we need some more groundwork in ALSA done. Due to lack of manpower I unfortunately don’t see this coming any time soon »
Also note, that you can always drop to “alsamixer -c0″ to fiddle with all the exotic mixer controls directly. It is very unfortunate that this might be necessary for now, »
Yes this is very unfortunate that you have to tell people to use alsamixer and that won’t change anytime soon.
Actually, nor I nor any of my friends have problems with pulseaudio - and we're quite a few (however statistically insignificant).<br><br>I play music all day and have quite a few simultaneous sound events. Had my first crash yesterday after a week of uptime (simply fixed by a new pulseaudio-session).<br><br>Pulseaudio fixes one MAJOR problem for me - switching to USB headphones (necessary for work). Also makes handling sound in vmware a simpler issue (no more glitches).<br><br>It is sad to see so many people having problems. It feels a bit like it is a shame that this last troublesome area of linux use is not a top priority for all the distros - they're all so good and this is the last beauty spot (at least of the caliber that it makes you nauseous).<br><br>Well.. what do I care - I am happy ;-)
The misconception is that pre-PulseAudio we never had a "working solution" as you call it.<br><br>Laptop usage is a perfect example. Mine also has external speakers and a build-in mic. Now what if I connected my usb headset? It was a pita to set up with pure alsa and nearly required manual tweaks for every application I wanted to use it with. With PulseAudio and the nice integration in 2.25.x it just works and even my mother could switch inputs and outputs now.
@pel: I think a lot of people don't actually have any problems with PA if they would just give it a try in the next round of distributions. It has much improved and now it is up to the distributions to ship everything correctly preconfigured etc. It's the right time to move on. New technology will never work for 100% of the users but if it is still possible to just remove pulse most users will not even care to file bugs etc
I don't own a usb handset or external mic and don't plan to. I don't need to switch input and have never needed. Am I the only one ?
I DO have problems with Pulseaudio, and a lot of other users too. Just look for bugs like this:<br>https://bugs.launchpad.net/bugs/288093<br>it's not that hard to spot them.<br><br>I, for example, have continuous problems with Firefox+Flash, Skype and other applications. It practice it means that if I have watched a flash video with sound then I won't be able to use Skype and vice versa. This is painful in 2009 with recent distribution like Ubuntu 8.10. I don't think that blaming proprietary non open source applications for incorrect usage of sound device is the right answer because I don't remember such problems when we had only Alsa
I only wish there could be a clearer discussion of the technical issues and presentation of the possible solutions. PA seems like a crazy solution but maybe it is necessary if only the facts could be laid out clearly.<br><br>Personally, I never had problems before pulse-audio. As for sending different sound to different devices there seem to be simpler ways to achieve that (I don't want to start another argument by mentioning Phonon).<br><br>I know little to nothing about linux audio plumbing but extending ALSA seems more sane to me than PulseAudio. There may be issues with that but are they as major as the issues associated with "faking" ALSA, xine etc for the purpose of rerouteing your audio?<br><br>It's not just a matter of an immature system, it's a matter of a very messy (even if it is a very well designed sort of a mess) system in which it is hard to track down the gremlins.
I didn't realise anybody else had to do "alsamixer -c0". I stumbled upon it trying to find a way to make the line out on the front of a Dell Inspiron 1525 laptop work under Fedora 10.<br><br>And yes, there ARE seemingly random issues with pulseaudio like no sound suddenly and unrecoverable without a full reboot.
@karol: I'm sure there are people that have problems. I'm just not sure that they're all that many.<br><br>From the people around me I see two loud people complaining. One does it from a embeded systems perspective and one has an old fedora installation. No one else complains (at least in my little e-bubble of people i usually communicate with).<br><br>I am not saying that there are no problems - I am however wondering if they're only a vocal minority?<br><br>@Michael: I hope you're right :)<br><br>@everybody:<br><br>Sorry to hear you all have problems - I really hope that they get fixed. Sound on linux has sucked arse for years and it would be a shame if pulseaudio was hosed down when it actually improved life for a lot of people.
maninalift: your comment evidences your lack of depth in understanding the problem. PulseAudio does not FAKE ALSA. There is MORE to audio than just kernel-level API -- mixing, routing, and real-time requirements are unfulfillable using either OSS or ALSA, and pulseaudio fixes those. The only problem here is BADLY PROGRAMMED CLIENT APPS. That is it. In strict rigour, it is the apps that need to be fixed, NOT the sound server. And in all practicality, abandoning something as useful and as fantastic as pulseaudio is the WRONG solution to the problem.
I also still use plain alsa under my Gentoo system. I don't need pulseaudio for now, like I don't need compiz and I don't use them.<br><br>Anyway, I am sure that Lennart and other cols like Colin are doing a great work. On the other hand, there are people that usually blame on distributions not providing proper setup. This could be true but, if distributions need to make so much changes in defaults upstream configuration, maybe upstream could modify some of current defaults for hopefully getting a workable pulseaudio setup without having to apply a lot of patches. Look for example to mandriva pulse package:<br>ftp://ftp.free.fr/pub/Distributions_Linux/MandrivaLinux/devel/cooker/SRPMS/main/release/pulseaudio-0.9.14-0.20090109.1mdv2009.1.src.rpm<br><br>It applies some interesting changes like use speex-float-0 as default resampler instead of speex-float-3 (current upstream default) as it is a better compromise between hardware resources and sound quality. Why is this not chosen as default directly by upstream? This would prevent that ubuntu, opensuse, gentoo, fedora, mandriva... need to change it
Rudd-O: First I fail to understand what you mean. You say that pulseaudio can fulfill real-time requirements that ALSA can't ? I really don't see how this would be possible.<br>For the second part, maybe the apps need to be fixed, but then pulseaudio should not be shipped in released distros until they are. Having a theorically great solution that breaks thing for them is of no use for the users. <br>The issue is that current choices make it impossible to not use pulseaudio when you are a user who can't fix the app.
A long term goal of pulse is for there to be a single UI for controlling your audio. There should not be different audio UIs in every app that makes sound. Get rid of those UIs and use the one in Pulse. For example, should every app have a different UI for switching between speakers and headphone, volume control, equalization, etc?<br><br>This is a historical problem in the design of Linux that needs to be fixed.<br><br>I agree that many of the problems people are blaming Pulse for are actually problems in ALSA that need to be worked on.
Jon: I don't oppose to a long term switch to pulse, or even to try having most people using it now. I'm opposed to forcing people who need their machine to work into using it right now given the know issues in alsa and various apps.
@Pacho<br><br>I readily admit that I have a lack of understanding of the problem. Indeed, that's really what I'm saying: users like me think (1) "I didn't have any problems before" and (2) "ok so there is a problem but surely there is a more organic solution"<br><br>Yes the people working on PA understand the problems far better than me, but I'd just like to be convinced. That's all.<br><br>You don't like the work "faking", well "mimicking the ABI" then.
As long as everyone's making platform-wide recommendations based on trivial anecdotal evidence, let's throw this one in there as well:<br><br>One laptop: ALSA horribly borked, first no audio at all, two months and 6 ALSA patch revisions later stereo speaker output, but no headphone output and no microphone input. Month afterwards, headphone output can be manually toggled on and off, and speakers muted. One year later, microphone input starts to work, though with amazing amount of static. At first could not suspend/resume while sound modules were loaded and anything in /dev/snd/* open (which would be all the time, effectively, either due to mixer app or an open web page with a Flash component on it). Can't see what Pulseaudio could do to make matters any WORSE, and in fact, pactl suspend-sink would allow the machine to suspend while direct ALSA access would have prevented that. Score 1 for PA.<br><br>One HTPC system with analog 5.1 audio: Again, ALSA horribly borked, with channels in wrong order, center channel coming out of the left surround speaker, and subwoofer repeating surround channel. Mplayer does one thing, Totem another, and MythTV third. Impossible to find a working configuration, and every application needs to handle stereo-to-5.1 upmixing on its own. Pulseaudio magically reorders channels correctly with no configuration, and upmixes stereo while passing through multichannel audio transparently with zero configuration. Score 100 for PA.<br><br>Score until now: ALSA 0, PA 101.<br><br>Broken apps: Audacity can't deal with a sensible sound API, demands direct access to hardware and exposes a mess of controls. Skype hides all controls but also demands direct access to hardware. And so forth. Score: ALSA 1 (by default), applications -100. Thanks to people who keep fixing these!<br><br>Final score: AlSA 1, PA 101, broken applications -100. PA wins. Thank you for playing.<br><br>User-space audio management and dumb drivers is the only way to make things work correctly. Every other operating system got this right 10 years ago, and until Lennart heroically introduced PA, Linux was a bog of broken drivers and hack solutions. It still is, but at least now that ugliness can be covered under something that makes sense.
Last time I checked, which would be last week, pulseaudio is easy to disable if you want to.<br><br>chmod -x /usr/bin/pulseaudio<br>mv /etc/asound.conf /etc/asound.bad<br><br>Log out.<br>Log in.<br><br>*POOF* <br><br>No more pulseaudio.<br>--------------------------<br><br><br>And the problem is before pulseaudio Linux audio was NOT WORKING for anywhere near 99%. <br><br>I don't think it's 80% or even 70%. I think the number is probably closer to 40%. The way things were were a chronic issue with Linux desktop... turning cheap sound cards that worked perfectly well with other operating systems into barely usable and seemingly flaky POS that does not work with many games and has a bout a third of the working features that it does in Windows.<br><br>If your telling me that the Linux sound system works for 99% of the people then your completely and utterly delusional about the state of Linux audio. <br><br>It's BROKEN NOW. It was broken before. PA solves a lot of issues, but breaks it in other ways. But PA's problems can be solved and it's necessary.<br><br>Just because something worked for you doesn't mean that it works for other people.<br><br>---------------------------<br><br>Here is a typical use for audio that you may not be aware of:<br><br>I bought my brother a web cam for X-mass. It has a built in microphone.<br><br>When you plug it in it is detected by Alsa as a usb sound card with recording only.<br><br>As you probably know Linux bases the order of it's devices on first come first serve. <br><br>So he shuts his computer down at night. <br><br>If he has the web cam plugged in when he boots up it becomes "Card 0". And his onboard audio becomes "Card 1". <br><br>If he boots up without the web cam plugged in then the onboard becomes "Card 0" and then when he plugs in the web cam it becomes "Card 1". <br><br>This means that, without PA, programs will seemingly work or not work. It was completely mysterious to him and he had no idea that is sound worked or not by what USB devices he had plugged in at boot-up. <br><br>And what made it worse is that removing the USB device didn't make things work. <br><br>Later on he figured out that if he selected 'safe settings' in his BIOS then that caused the sound to start working.<br><br>However his mouse then stopped working.<br><br>This is because his mosue is USB and safe settings in the BIOS disabled the USB devices on the front of the computer.<br><br>So he calls me up saying he couldn't get any sound in his games until he played with his BIOS, and now sound works but his mouse doesn't.<br><br>He is basically computer illiterate. Knows just enough to be dangerous.<br><br>It took me 4 hours on the phone working with him to untangle this cluster fuck. <br><br>This is f****ed. It's just that bad, that confusing. <br><br>Linux works well for stereo playback of music files or video files.<br><br>THAT IS IT. It's too confusing, too convoluted, too easy to break to be useful for anything much beyond that unless you spend a great deal of time studying how Linux audio works.<br><br>With PA I can get it to work irregardless of what USB devices he has plugged in at bootup. <br><br>Oh, and he also has headphones with a Mic that he wants to use. <br><br>Now you tell me what sort of magical asoundrc file do I need to whip up to get both his webcam mic, and the headphone mic working, be able to switch between them, have applications work with both of them, and then have his audio work irregardless of which device shows up as 'Card0' and Card1' (or what udev rules I need to create to keep the devices in order).<br><br>I can get this working sorta through PA. I'd need a master's degree to get it working in any other fashion.<br><br>Then also think about how my 'USB Dock' with it's USB audio is suppose to work.<br><br>This is not atypical stuff. People want to be able to plug in devices into Linux and have them work. They want to be able to make recordings, plug in their web cams, plug in their handy cams, use VoIP, use all that stuff.<br><br>Right now the amount of people that gets that 'working out of the box' is very very very low. I know it's never worked for me without spending a great deal of time farting around with it.
I have been trying to get pulseaudio working with my gentoo amd64 system for the last week. In my case, one of the requirements is that I use Jack for some of my software. <br><br>Pulse comes with source and sink modules for jack.. Once it is running under Jack, I can make pulse segfault by switching songs under banshee. I am trying to gather enough information to file a bug report.. for now, as a workaround, starting jack runs a pulseaudio -k and then restarts it when the jack daemon stops. <br>There seems to be a lack of documentation for what I would think is fairly normal use case as a lot of audio software is designed for jack.
osma: Audacity doesn't actually "demand direct access to the hardware", it uses Portaudio, which until recently could not work with Pulse because it didn't support non-mmap access to ALSA. A Fedora contributor quite recently patched Portaudio to work non-mmap (and hence work with Pulse). The patched Portaudio is available in official updates for Fedora and in /main/testing for Mandriva 2009 (because somehow it doesn't work for the QA team, despite working fine for everyone else who's tried it).
Nate: "and then have his audio work irregardless of which device shows up as 'Card0' and Card1' (or what udev rules I need to create to keep the devices in order)."<br><br>No udev rules. What you need is this, in /etc/modprobe.conf<br><br>options snd-(card-driver) index=0<br>options snd-usb-audio index=1<br><br>where snd-(card-driver) is the actual module name for the internal sound card. If you do that, the internal card will always be the first device, and the USB mic will always be the second.<br><br>Not a lot of people know that, though. :)
Ya. Thank you. I'll have to remember this.<br><br>It's all just a bit infuriating. The biggest problem right now is just getting people to admit that there is a problem and getting people to look at the issues objectively.<br><br>It's like when everybody was using Windows 98 for 5 years or more they've all conditioned themselves to carefully avoid doing things that broke that OS. As long as you were experienced and knew what to avoid that was actually a fairly stable OS that only needed to be rebooted once every few days for memory leaks. <br><br>But it was still a pile of shit. Convincing people of this after they've been so well self-conditioned was still extremely difficult, unfortunately.<br><br>It's kinda like that for Linux audio. People get set in their ways and simply avoid doing things that cause themselves problems.
I personly switch off PA on MDV2009.0 and 2008.1. I need a softphones (tried ekiga and twinkle) to work and after a day of testing I found out, that disabling PA and using ALSA is easiest way for me to configure sound system for softphones. Then you want to start testing your asterisk server and you cannot do it because your mic is not working it is starting to ... and then you go to other machine and then again the same problems.
I have two different PC's running Ubuntu 8.10 (fresh install). Both have problems with Pulseaudio. The main problem were that all of the sudden audio (and video) stop for a second and continues (with a short skip back). Very very irritating. Uninstalling (+purge) pulseaudio fixed all those issues.<br><br>Before pulseaudio I never had any real problems with Linuxaudio-output on a lot of different machines (incl laptops). Only input needs some configuring (e.g. enabling capture). Maybe there where some technical issues, but I never really noticed it.<br><br>Hopefullt the pulseaudio-issues will be fixex in Ubuntu 9.4.
Nate: oh, and I missed something. 'irregardless' is not a word. If it were, it would mean something like "regarding", because it contains an internal double negative (the 'ir' prefix and 'less' suffix are both negative in effect). The word you're looking for is either 'regardless' or 'irrespective'.
Michael,<br>> it is up to the distributions to ship everything correctly preconfigured etc<br><br>The real problem is that in some cases some dumb programs simply cannot be correctly preconfigured, because they are dumb. E.g., show me how to configure Ekiga to use one sound card for the speaker+microphone and the other one for rigntones, both through PulseAudio. Also (although it is not a PulseAudio problem) how to make it ignore my TV tuner as a video source (I don't have a webcam).
Glad to see I'm not the only person who had PulseAudio fix their setup. To allow more than one application to output sound at once, I have to use dmix with ALSA - which causes horrendous audio distortion when above 5% volume. When I'm using PA, it works perfectly.
Alexander, <br><br>Ya. I have had problems with Ekiga and PA. I can get it to output fine over PA, but input from the microphone is a problem. It's a bit odd because lots of other programs can record perfectly well over PA, but Ekiga is just screwed up somewhere. <br><br>I am hoping that the telepathy stuff works out. As much as I want to like Ekiga, it's just not a pleasant program use. Combining IM with VoIP is just the way to go, IMO. That other people get to know when your available to talk and sometimes it's just easy to type stuff out then say it. Often it's also more convenient to type stuff out because often you only have something simple to say and you don't want a entire conversation over it. <br><br>Text + VoIP + Video + just seems natural to go together.<br><br>-------------<br><br>Adam, <br><br>Sometimes some things just don't really matter that much. I like that word even if it's not a word. It amuses me. :)
You can also use KMix if you want a low-level ALSA mixer. AFAIK it works in GNOME too. Maybe other environments, e.g. XFCE, also have a mixer tool you can use.
Hello. And Bye.
Hello. And Bye.
Hello. And Bye.