Gentoo EAPI 5
This is a modal window.
Das Video konnte nicht geladen werden, da entweder ein Server- oder Netzwerkfehler auftrat oder das Format nicht unterstützt wird.
Formale Metadaten
| Titel |
| |
| Serientitel | ||
| Anzahl der Teile | 84 | |
| Autor | ||
| Lizenz | CC-Namensnennung 2.0 Belgien: Sie dürfen das Werk bzw. den Inhalt zu jedem legalen Zweck nutzen, verändern und in unveränderter oder veränderter Form vervielfältigen, verbreiten und öffentlich zugänglich machen, sofern Sie den Namen des Autors/Rechteinhabers in der von ihm festgelegten Weise nennen. | |
| Identifikatoren | 10.5446/40018 (DOI) | |
| Herausgeber | ||
| Erscheinungsjahr | ||
| Sprache | ||
| Produktionsjahr | 2012 |
Inhaltliche Metadaten
| Fachgebiet | ||
| Genre | ||
| Abstract |
|
FOSDEM 201251 / 84
1
2
3
5
7
9
11
13
14
15
16
17
18
20
23
24
26
27
28
30
32
37
39
41
44
46
47
50
51
52
53
55
57
58
63
68
70
73
75
79
80
81
84
00:00
GruppenoperationEntscheidungstheorieWhiteboardInformationDateiformatVorlesung/Konferenz
01:10
InformationVektorpotenzialInhalt <Mathematik>DateiformatInformationPunktRechenschieberComputeranimationVorlesung/Konferenz
01:56
Produkt <Mathematik>SchaltnetzFahne <Mathematik>BinärcodePhysikalisches SystemGleitendes MittelQuaderSchnittmengeProgrammfehlerURLBinärdatenVorlesung/Konferenz
04:51
SteuerwerkInhalt <Mathematik>ProgrammierungInterface <Schaltung>MultiplikationsoperatorPhysikalisches SystemQuaderCASE <Informatik>DatenverwaltungHardwareHumanoider RoboterClientKonfiguration <Informatik>MultifunktionDefaultPunktBinärcodeComputerarchitekturMAPARM <Computerarchitektur>ImplementierungRechter WinkelCookie <Internet>BinärdatenVorlesung/Konferenz
07:11
VersionsverwaltungMAPMultiplikationsoperatorMomentenproblemZellularer AutomatFrequenzData MiningGeschlecht <Mathematik>BinärcodeGebäude <Mathematik>ForcingBasis <Mathematik>AggregatzustandProzess <Informatik>Keller <Informatik>SoftwaretestElektronischer FingerabdruckMereologieProjektive EbeneInverser LimesDistributionenraumBitPhysikalisches SystemServerTypentheorieSchnittmengePunktGruppenoperationRechter WinkelDokumentenserverWechselsprungBinärdatenKonfigurationsraumDatensatzReelle ZahlVorlesung/Konferenz
13:25
FaktorenanalyseBitDatenverwaltungStabZahlenbereichMultiplikationsoperatorImplementierungVorlesung/Konferenz
14:12
FaktorenanalyseProdukt <Mathematik>VektorrechnungMultiplikationsoperatorGraphfärbungElektronische PublikationZahlenbereichAbstimmung <Frequenz>Projektive EbeneBenutzerbeteiligungQuaderZusammenhängender GraphProdukt <Mathematik>MathematikComputeranimation
15:28
VariableEinfache GenauigkeitFunktion <Mathematik>MultiplikationsoperatorVariableUmwandlungsenthalpieLaufzeitfehlerDatenverwaltungGüte der AnpassungMeta-TagMAPKreisflächeVorlesung/KonferenzComputeranimation
17:11
PhasenumwandlungFunktion <Mathematik>QuellcodeCodeRechenwerkMetadatenElektronische PublikationSkriptspracheSkalierbarkeitKonfigurationsraumBinärcodeMetropolitan area networkDeskriptive StatistikServerMetadatenRhombus <Mathematik>FunktionalDefaultHalbleiterspeicherBildschirmsymbolPerspektiveDienst <Informatik>Klasse <Mathematik>SoftwarewartungWürfelMultiplikationsoperatorDiskrete UntergruppeVersionsverwaltungCASE <Informatik>ParametersystemFahne <Mathematik>AppletPersönliche IdentifikationsnummerFormale SprachePhasenumwandlungUmwandlungsenthalpieRauschenHumanoider RoboterSchnelltasteE-MailOrdnung <Mathematik>QuellcodeVerzeichnisdienstGruppenoperationKonfiguration <Informatik>MAPSchnittmengeStreaming <Kommunikationstechnik>Kartesische KoordinatenVererbungshierarchieMailing-ListeRechter WinkelDatenverwaltungObjekt <Kategorie>BinärdatenBitGebäude <Mathematik>Offene MengeProgrammfehlerOffice-PaketArithmetisches MittelMereologieComputeranimation
25:23
MetadatenMathematikBitSoftwareDeskriptive StatistikInhalt <Mathematik>PunktComputeranimationVorlesung/Konferenz
26:17
BitElektronische PublikationArithmetisches MittelDeskriptive StatistikMetadatenSpeicherabzugSkriptspracheInformationDatenverwaltungSchnittmengeAdditionMathematikSynchronisierungPunktHomepageNumerische TaxonomieGerichteter GraphVisualisierungKonditionszahlÄhnlichkeitsgeometrieVorlesung/Konferenz
29:54
HomepageMetadatenMathematische LogikKlasse <Mathematik>InformationGrundsätze ordnungsmäßiger DatenverarbeitungKonfiguration <Informatik>MereologiePunktMaßerweiterungMetropolitan area networkAuswahlaxiomPackprogrammDatenfeldElektronische PublikationQuellcodeBitCASE <Informatik>DatenverwaltungZahlenbereichMetadatenObjekt <Kategorie>InternetworkingSichtenkonzeptE-MailWurzel <Mathematik>TopologieSoftwareentwicklerVersionsverwaltungRepository <Informatik>ComputeranimationVorlesung/Konferenz
36:39
InformationDifferenteZellularer AutomatElektronische PublikationPunktMultiplikationsoperatorMaßerweiterungVariableDefaultE-MailPackprogrammSoftwareentwicklerCASE <Informatik>Vorlesung/Konferenz
38:03
Mathematische LogikKlasse <Mathematik>ZählenDatenverwaltungHook <Programmierung>TopologieVersionsverwaltungInelastischer StoßServerQuellcodeElektronische PublikationMultiplikationsoperatorVariableCASE <Informatik>KollisionserkennungQuick-SortGebäude <Mathematik>DokumentenserverGeschlecht <Mathematik>PackprogrammTreiber <Programm>Fahne <Mathematik>SichtenkonzeptSpannungsmessung <Mechanik>DifferenteVorzeichen <Mathematik>MultiplikationDatenfeldKreisbogenStellenringMehrrechnersystemSoftwareentwicklerZellularer AutomatKonditionszahlp-BlockWort <Informatik>Computeranimation
45:22
Elektronische PublikationInstantiierungMultiplikationsoperatorGebäude <Mathematik>Physikalisches SystemTopologieTeilmengeBinärcodeMereologieDämpfungFlächeninhaltDateiverwaltungVorlesung/Konferenz
46:51
Notepad-ComputerGruppenkeimDatenanalyseElektronische PublikationDateiverwaltungInelastischer StoßRoutingGebäude <Mathematik>CASE <Informatik>VollständigkeitStellenringBinärcodeSchnittmengeTermE-MailAdditionComputeranimation
48:29
Lokales MinimumElektronische PublikationVorzeichen <Mathematik>Überlagerung <Mathematik>Virtuelle MaschineComputeranimationVorlesung/Konferenz
49:16
CodeExpertensystemMigration <Informatik>Just-in-Time-CompilerVorzeichen <Mathematik>HypermediaDämpfungServerElektronische PublikationProzess <Informatik>ProgrammfehlerHinterlegungsverfahren <Kryptologie>MultiplikationsoperatorSchlüsselverwaltungSynchronisierungGeschlecht <Mathematik>BootenBitSummierbarkeitTopologieNummernsystemBenutzerbeteiligungPunktSymmetrieGerichteter GraphMAPInstallation <Informatik>RelativitätstheorieDrahtloses lokales NetzWort <Informatik>FreewareVorlesung/KonferenzComputeranimationBesprechung/Interview
Transkript: Englisch(automatisch erzeugt)
00:00
So the title of this talk is Gen 2 APA 5 and you can alias it as Gen 2 bike shedding or whatever so Who am I? My name is Peter. I'm a Gen 2 dev since 05 Council DevRel recruiting so basically
00:22
sitting in the board that makes technical decisions in Gen 2 so the idea behind this talk was to A-collect devs for the dinner afterwards, so I don't know let's see how many devs we have now Largest not least one in there There we have a group there some more so quite a few
00:45
Good other Andreas wasn't listening so but they'll fridge so the question was how many Gen 2 does we have in the room and we had
01:01
Hands popping up around the room So what I did for the talk was Added some basic information on the API how many of you know e-bills the the file format most men at all so
01:22
So e-bills is the packaging format for Gen 2 and feel free to ask questions at any point Because some of the items probably need more these more background What I have on the slide, so I'll explain them more yeah
01:56
The question was will there be any way to have official bin hosts for
02:02
Gen 2 packages There there actually already is binary packages in the tinderbox dot Who knows the exact role tinderbox something dot gen 2.org that you can use to rescue your system, but?
02:26
So people want SSL on it Yeah, I'm repeating the question that the question was could we get SSL on the tinderbox URL You you that just have you opened a bug request
02:49
They have the open a box to the gen 2 infra product or I think that that it that it will do it Then I'm quite sure they'll be happy to add it. You know they're already and of course
03:06
Do we have Fabio here kuleani He does More binary stuff from portage that we don't know more, but I don't know infrastructure blessed that the problem with binary package is that
03:28
Unless you want to build for all the use flag combinations. You're always gonna be tied to a single pre-selected set So do we have the other other mic sorry?
04:20
Sorry is the other mic working Okay, so then we can get that that discussion recorded live So the I think the summary is that if you take away use flags and compile the binaries. Why aren't you using Debian?
05:07
The question was Was to have a point in the middle because he builds are great They are far better from specs and and devs by far. We have several options that you cannot
05:21
Implement in any way on other package management system, but the thing that is maybe maybe refraining users is is the time of compiling and the big time in global so maybe if having having Something quite intermediate between using binaries that are that corresponding 90% of the case
05:45
to the to the specific hardware that is On the client side Plus plus having the right to compile and Feeling free to compile that doesn't work with Android and with with the ARM architecture of course I
06:02
Think if you have any people here who do releases I think the release is used used to come with binary packages that you could install. Yes, of course, but I mean having having by default Portage bin hosts For almost Almost all almost all the cases meaning. I don't know you have people who have an
06:24
ATI Nvidia Intel architecture and AMD that makes quite a lot of simplification and that's can That can save a lot of compile time Is it on the roadmap of Gen 2 or I think it's a resource intensive effort besides
06:43
Automatic cinder boxing so I don't think I I'm not aware of anyone actually working on it into a big cookie unless anyone there has a You have a comment You're working on it, so we have when will it be ready
07:06
Okay There another comment. Yeah Actually, I think instead of having some common set of binary packages
07:21
It will be much more useful to have the easy way to configure Bin hosts for periodic rebuilds and have custom stage force with binary packages and and so on and I'm working on this as a side project because I I want it and I need it
07:45
and I have run into a few problems and with this testing package is not really building and yeah, but I think This is really not
08:00
The issue is not about having official Gen 2 packages, but making making some binary packages Available there was with Linux. Maybe it's still alive. There is someone if you want desktop desktop destroy easily
08:20
installed easily and have Gen 2 which or at least you can Use it as a as a Gen 2 bootstrapper, but it's true. There is nothing of Such a type for servers and I think it's a bit lacking for easy installation of servers or
08:44
And anything on on the server side, so I think a tool For easy easy building of the stage force and binary packages. I I've tried catalyst. I've tried metro and It didn't really get the job done well for me
09:02
Because I want really especially tailored state stage force and they didn't step out out of the out of the Limitations of the Basic release, but I'm going off topic with this. Yeah
09:24
So, but the idea is to have good tool for Creating been hosts Fabia arrived. There was a question. I will Jen to have officially blessed binary hosts I don't know if you have anything to say on the topic because you're interested in binary packages more than I am
09:45
That there's it. I Didn't hear the question from the beginning. So Could you try to summarize it? There was a request. Will we have a blessed binary? A package repositories in Gen 2 if we have it. Oh, no, though. Is it on the road map? Oh
10:09
You've done some work on binary packages, but yeah, I know I don't have rebel wise but point is wise I Actually didn't get the question, right
10:22
So what's the actual question the question is from there from next year, but yeah On the same row there now this guy my main purpose was to Have more it's a more newbies come to gentle
10:45
That was why I was asking the question and why I was proposing it and by the way, I find that These guys idea is is better than mine because you can select at which stage you can have a package instead And that can be a very great
11:02
So that the only the only purpose of this question was to have to have the community come back to Gentsu and especially newbies who wanted to Even if we don't want it
11:22
Well, I Think that a good path to to come to Gentoo is actually try to use Sabaeon Which is the easy part the easy version of Gentoo, which is its purpose is actually to catch newbies catch users from Ubuntu or arch or whatever distro and and
11:44
yeah catch and bring them to a gentle basis system and Actually the next step would be for the user if one wants to know more how Our system can can be configured etc could easily from Sabaeon
12:04
Get to a real Gentoo system actually, so The path for newbies I think in my opinion is to try Sabaeon distribution Which is completely 100% compatible with Gentoo
12:22
Stick with it for a while Know how the partage system works etc, and then make the jump if one wants I think that's the The way to do it in my opinion and with Sabaeon you get binary packages So if you just want to get a system up and running you just need to install Sabaeon and be happy
12:48
Okay Did I answer your question? Okay There's there's always on the with newbies and gender. There's always the question that Genju doesn't want to attract the
13:01
first time Linux users on there's unless they're really involved in getting to know as it's they they don't get From other Genju what they probably would have gotten from Ubuntu as I think if you if you see it as ISO z-stack Gentoo is a level 3 and we are living for so
13:25
if you just want to go down and implement your stuff just go down level 3 and and Have fun with UDP or whatever So coming back to the topic of EAP is the history wise be being cranking them out
13:43
Pretty much one per year, but I don't know what It is time for the next one, so that's why I want to do the talk and EAP one I don't know that your number it wasn't in the
14:01
Tag, so but sometime before that So why is it a bit slow Portage the main package manager is a bit takes a while to implement new features and and of course you can take time to Pick the color for the shed For example a long discussion about should we add the EAP I the
14:24
Number to the file name at the end Was going for quite a while and never it went quite close to in the end of the tie votes Since then it's been pending
14:41
so two actual features what I did I Went through the in the box to attend you to Gorg under the product change you hosted projects and the competent PMS Slash API you can find Ideas floating around to be implemented in API. So what I did was go through the list and pick some
15:07
Interesting ideas are things that I would find useful to be on the slides And so we can discuss and all the devs who don't agree with the ideas can shoot them down and then at the end if you have any more ideas besides these then we can
15:25
Discuss those from the audience Now what I would personally like the most is Exurbo I think does this already but To read get rid of the currently there are two variables build time deep actually three of the two main variables build time depend and
15:45
already bent for runtime dependencies and then special circle circle dependency breaker called key depend post depends So what the discussion is that you have them in a single variable and in the more
16:02
defined way currently If you look at the spec package manager specification on how the dependents are defined it's very It's not really clear because of the historical way and portage work, too I would want to separate them to a more criminal level and
16:25
Specified in that way, but I think there are people who object this meta Of it one of those people in the audience No, not today good so I can say that everyone here supported that idea
16:42
So what what happens currently is that with binary packages use are already bent but if you have some for example what a Benefit I could think of is that if you use something only in package post install you would have to Put it to our depend because the binary packages are running that step
17:04
But even though you know needed a runtime export functions that anyone any people here who have written a classes
17:22
So Do you do you agree with the assessment that export functions is totally unnecessary It is a way to tell a list in the file that which functions the a class Has so that could be just basically automatic and removed no objections good
17:47
so when writing so What happens when you use e-bills is that their steps? basically if you the traditional configure make make install, there's a free for e-bills there's
18:04
matching steps and bit more unpack prepare configure compile install For the source So in the e-classes you can provide defaults and Then if you don't want to use those eat defaults in the e-bills you have the you can
18:24
specify your own But truly there's no way in the e-class to sell that I want this at source Configure time I want this e-class function to be always run Some people make the
18:41
Argument that e-bills themselves should then Just be calling that e-class function, but at least with Java classes and those that have Needs with this find that I find that it's very hard to remember to do it and crying hard to automatically catch Let's set up a lot of infrastructure. So why not have the package manager take care of it?
19:14
That is well at least my common use case doesn't matter because there's they tend to be
19:20
more like checks but That's the time for bike shedding And I'm just gonna get the mic to pick a worst case if you have is if you have an e-built for language bindings
19:46
There's a KDE package I'm You want to call the class for the language, which is something like Java e-class or whatever And you also want to use the KDE classes and then you end up with
20:01
Wanting to call well kd4 base package setup and Java Package setup and then you don't know in which order and then you don't know how they interact with it with each other So and this needs some careful preparation, but it's a it's a definitely a very good idea that they are called automatically somehow It just needs to have a clear
20:23
It could get a bit that you specify you if you want you could have both free and after But The need is to at least the my this whole problem I'm sorry trying to solve here is that there's some QA and those kind of actions that
20:40
That they are not independent on the order. So and for that bad reason for that that use doesn't matter which are there For the general solution. Yeah, true Are all good up streams all the upstreams really source code and it's always easy to package right in that but
21:10
There's the discussion that if you want to if you need to create The D file that's get distributed on the servers What currently happens that people maintain their own scripts somewhere?
21:24
for example we provide binary packages for open JDK just because it takes a lot of memory and time to compile and Libra office too and things like that, but so So To package the you could for example use this to have the steps to create the binary package
21:47
For those special occasions then in the in the bug you find that this question about more generally starting to have more mean maintenance or maintainer only
22:01
Phases in e-bills It was Spanky or who was suggesting it but that we could move QA checks to maintain a specific functions To reduce the noise from users that that's probably controversial but That is also a part of the discussion in that
22:24
So and any devs here who would have used for this this kind of stuff So the best application is auction oxygen icons Because we always repack it because because nobody in Gentoo has ever used for the scalable SVG icons
22:42
So I mean the tar file that you download from from the KDE servers somewhere around 350 megabytes or so and we end up with a 25 megabyte file. Have you but upstream can strip it down? Upstream can but refuses to do because they say it's a licensing issue
23:10
We only strip it down if it's not with the use flag bin package set
23:34
So next item I found was to move description To metadata XML there in any build package directories. You have these XML file
23:46
Having package level metadata and then you have the individual e-bill files that have a e-billed specific metadata Which includes the variable description? so Fabio might know but I'm not too sure what Descript which version of the description port it uses when it tries to show you the package description. I'm assuming it might be the latest
24:09
But Yeah, the answer was he only asked the API and he doesn't know what doesn't do it
24:27
So at least from my perspective the package should have Only one description by default In a single place, but then then this comes to this discussion then
24:41
Is there use case for having a version specific? Described descriptions that we need to be shown at all Maybe or maybe slot specific if there's a new major version. Ah Diego
25:00
You could get Andres. Can you pass the mic? There is some use case for that but we have a solution for what as well metadata has a restrict keyword But we can use to restrict the description to a particular version or a slot or a sub set of versions
25:22
So shouldn't be a problem. Yeah, the question was more should be Allow that at all or just have one description. We have used it before Especially between slots of the same package because sometimes The content of a package changes so drastically between slots, but you don't have the same software
25:46
So yeah We have solved already some time ago, but you don't need the description to install it though
26:12
Yeah, yeah, it's it's a good point I haven't I haven't considered that Well in general it is that a package is a set of e-bels
26:24
Ideally, of course, they don't live that long in the bugzilla but Fabio so we also have the long description Tag, I guess in metadata XML already. So what's the point in having?
26:41
One kind of description in metadata and one another kind of the script description in the e-bill doesn't make sense We do but of course we do need a short and a long one, yeah, sure sure sure that for sure But not in in different places actually. Yeah, I am for removing it
27:06
Yeah, but as far as time and for for Moving actually from for moving description to I agree. Yeah, so let's move
27:25
the long description to the e-bels Given I was the one proposing the addition of description to metadata from use description that's another thing that moved from
27:40
Another file they use local file to the metadata XML and I've been Pressing for this to happen for quite a bit metadata dot XML doesn't have only Core information they have a lot of user facing information like they use descriptions. So it makes sense to keep using that for user
28:03
user information as well and since it's XML the package manager can simply Handle it for the user and show it to a user transparently there shouldn't be any reason why the package manager and those the description in the e-bid file itself differently from the
28:23
data from the metadata so doesn't make sense to have to from the point of sight of what you commit and decreasing the size of a builder is also always a nice thing because Our sync is slow. Can you get the mic?
28:54
I'm gonna fall back to taxonomy because metadata means describe data. So
29:01
description So why only moving the description and not also the home page because it's similar
29:39
Assuming upstream home page doesn't
29:42
change every release Hopefully at least what will be it could but then again we can use that
30:04
Metadata options you have the home data version specific, but I would It wouldn't be my first choice my first choice was to would be the inment the famous device to stab people of the internet Why not true who's ever worked on an e-bill that has a zip file and forgot to depend on
30:38
I've arched on zip only a couple people. I mean
30:50
So one way or another I would like love to see a way to do that automatically I've seen Is do we have anyone here who has read the recent stuff on gen 2 dev the e-class that's been written for this
31:07
so Maybe you want to summarize and raise that
31:31
Just hope that whatever they're doing It's going to also support exotic stuff because it's always getting annoying when you have a 7-zip archive on ISO or whatever
31:59
Isn't it better to settle once and for all that the unpacker is either gzip or bzip2
32:06
Well tell that the upstream who distributes zips That was my humble point of view I would have no objections there, but I don't see it happening
32:24
We do have Packages that the source file is Rpm or dab expecting a zip file
33:07
The old JDK had the self-extracting file as well Yeah, but I very unpacked something first and then run that from the resulting file and then extract How would you implement?
33:28
The Detection of the unpacker during the the the the pre-configure of the e-bill well For the most cases I would imagine that if there's a file suffix
33:41
that if it's a dot zip it's not a Not a PC to file or whatever You can always fall back on doing things manually if it's a zip file that is actually a a 7-zip or whatever
34:00
Then you would just depend on that and unpack it yourself for the manual in source unpack But the common use case is that it's a tar.bc2 well that doesn't actually require anything, but dot zips are unpacked with unzip So you can map it based on the file name And if it doesn't work for you, then tell package manager manually what you need. I wouldn't rely on the file extension
34:40
Maybe you can include it in the e-bid If you rely on magic numbers to calculate the dependencies you would have to download all the files Of the whole of the whole dependency tree and take a look yeah So you would have generated dependency information coming with the tree or how how about just adding?
35:08
Kuwait check on reprimand is that that the text the source file and tries to Search through the dependencies if there are dependent missing dependencies for that
35:23
zip file there is there is actually At least washer check for that. I know if it was a repo mano P check or The polish tool that can eventually move removed Okay, so
35:47
That's what I was going to propose myself As a fallback if the extensions can be relied upon Usually the developer knows what so we can just override it with a specific format, but just about it
36:01
Yeah, the root of the problem is that developers aren't perfect and they forget to say that information What is the unpacker um what what why make it a mandatory field if it's
36:25
If it care if we can do without it So just to tell that in eastern rebuild you would have to start telling people that this needs to be unpacked with tar Well for the normal case
36:41
Just about trying to make it convenient and safe at the same time so the fault taken from the Archive extension and then over over rideable by the developer if it's not reliable should be covering most of the cases There's usually also the case that if you make the simplest e-bills work as best as possible then normal user are able to come up with
37:05
usable stuff them to sell for their own use if there's plenty of Magic variables to define and things like that makes it harder for them to write their own stuff
37:22
Do we need at some point to separate the archiver From the compressor. Well, it's already done. It's for zip files. It's different package that has the Compressor and different package that has the unpacker
37:41
Yeah, but I mean when you have twice at the same time Tar BZ Tar GZ for example Do we maybe we might need to have both of the information? To prevent developers from telling my Akiva is just or AR for example, so
38:04
But if it's a target BC to target easy, it's still the same you can you put in the e-build our archiver car compressor visit to and Those two feet could be mandatory
38:21
Should be mandatory at the time we did In e-bills in the source you are a variable which sell tells the file
38:42
Sorry, what if your developer says it's a TBZ and the TBZ too?
39:04
yeah, in e API 2 you can specify the Automatic renaming of file when it is downloaded from upstream servers To something else. So if there's an unknown A Suffix what your solution cannot be implemented by just renaming it to a
39:22
File convention that the package manager already understands and then we don't need to have additional Variables that are always mandatory for that purpose Yeah, so to summarize that he was saying that the e-class will handle 99% of the cases
40:12
So we'll and leave the rest to the e-bills to override So I just have one little feature request But I'm gonna write that also in the bug or in the discussion somewhere namely that we can hook into the automatics
40:24
somewhere in the e-build for Let's just say pathologic Pathological cases like a driver package where you download an ISO which contains a target zip which contains a couple of RPMs Yeah, it happens
40:47
Ginger dough buzzer lazy, so they don't like new manager the flags. It's not gonna be popular Unfortunately, we don't have a better dictator anymore. I have a different feature request
41:10
Regarding file collision handling Often I have the problem especially in the
41:20
unstable arc after successfully compiling a package Yes after after that I get a file collision detection and that prevents the installing of this Successfully wheeled package and so you have to clean the file collisions by hand
41:48
Can we implement a feature Into the ap5 Of Which yeah, which offers a
42:00
Switch like do you want to overwrite the file collisions? Yes or not or something like this to solve it Yeah For the collisions if they are files that are there but I have no other package owning them it has been switched to
42:27
Collision it's Yeah portage these days on words if something else doesn't own the files various two features one still stops make that count
42:42
make the count there is Yeah, two two different features one stops for any collision and the other only the collision is with another package Where there should be a blocker in that package, but at the time there isn't one
43:02
It's one of the bigger Downfalls of the sort of building from source approach is that no one maintains a central repository Or at least the placed repository of files installed by e-belts and as they change by use flag It's would be an interesting effort and since it takes about six weeks to reveal a wall to you
43:29
Have you by the way a side question Diego have you managed to build the whole tree Whole tree, no, it's about
43:41
11,000 builds 10,000 build fine over 10,000 started either blocking each one and then Colliding Blocking not not correctly correctly functioning. So no the whole tree
44:02
I have never been able to be in there are lots of local so we can actually install all the tree but Even trying to build all the tree is not feasible right now the whole tree So there is Fabio had something I've been able to build actually
44:25
13,000 packages from Portage So it's kind of fun when you have to Rebuild packages new
44:41
I think 15,000 but I didn't count the multiple versions of the bills So I just build for one one version for each slot. I Think there we are at 30 30,000 Files yeah
45:02
Klondike, did you have a comment? Can you get the mics? How can you try to build a complete three if there are packages that depend on other packages not being there Packages depending on something not in the main tree
45:23
Pluckers, yeah Well, isn't the approach that you need to unmerge something and then we continue with the other set as for the tinderbox But instance the packages doesn't just build them the tinderbox tries to unmerge the packages that are colliding unless they are part of the system or of the
45:46
subset of file of system packages of known system packages, but the tinderbox needs to run and It usually works fine. There are situations where the presence of a given package you can break lose everything I have the problem with RPM for instance and
46:05
There are problems with my adb versus my SQL It is tricky very tricky. And that's why the tinderbox will never be able to target the whole tree. Just a subset As it happens, I have all starts So I last time checked out
46:22
11,000 packages So I I had to do the stuff from scratch There will never be a whole tree installed on the system. There can be a whole tree package tree Because you can build binary packages for everything as long as you remove every other packages over time
46:43
But you can't install it all together This can be necessary for someone who wants Who needs to have a portage being hosts a private portage be the host with all the package we build
47:01
That's the I think that's one of the only case where it can use That's still feasible because you can build Binary packages of everything as long as you unmerge some of it block the dependencies But you can't install on the same file system everything on the same route. That's can be done
47:23
But you can be by collisions blockers and sometimes even just the presence of a header file that is the same name of the file inside the package and the house Yeah, there is tons of corner cases. So five minutes
47:51
Okay, cool so This this I picked up and what latest additions in the E API for
48:03
Is that there there are operators for for the set but for the practical use just have one of Question mark question mark for the crew to signal zero or one from Ulrich Probably not controversial I hope and so
48:34
Who has an ice cream machine? Good one so any
48:44
Important ideas are things that you would like to see that I didn't cover in here and If there are those just let's make sure they end up in bugzilla
49:01
Please can you tell me the status of? signing of manifest files Status of sunny manifest files who would have a good answer here You can sign them, but it's not mandatory and the collapse have been approved It's waiting for someone to write it some people
49:21
wrath pieces of code To have all the infrastructure Which hasn't yeah, can we get the mic coming back down? I'm not really the ultimate expert
49:53
but But I think this mainly depends on how Whatever JIT migration will ever happen will turn out
50:02
Or is coupled at least to that question Because the manifest scheme will change with that probably there will be also commit signing Just in the very newest JIT release there's no upstream support of commit signing
50:21
So it's not it's not push signing, but commit signing I Guess that will influence whatever happens to the manifest signing as well, and it's not really decided yet At least that's what I know And also you can already get signed trees from infrastructure The snapshots are signed and they come from see CVS, and that's all then they get it
50:44
So if you want to get a sign tree, it's already possible. Just not don't use our sink It's very nice to have signed snapshots, but it's
51:03
Something Troublesome because there's no way to Or really check check it on the On the on the new install because there's no GPG on the stage free. Yeah, that is annoying
51:23
But then again it comes a question of how do you verify the install media Yeah, but if you trust the web server why don't you trust the keys that for the snapshots then so yeah the solution
51:57
Yeah, must have been outside Is there I guess there is no
52:03
GPG on the minimal media either is there There is yeah, but it's signed, but does it have the keys on the media it might be nice to have
52:39
a release or bootable media with pre-installed. Yeah, file a bug to release
52:58
From what I understand we are only based on MD5 sums
53:02
No, the if that was there is no signing process. I mean from scratch from the CD install to the effective deployment of the packages Before having to install GPG during the install process, can you get the mic so I
53:32
Relent should have the AHA Sam as well as the MD5 on everything so we are not limited to MD5
53:41
But no varies there as far as I can tell there isn't a new PG on either the install media or the stage But should probably be fixed when we implement the full symmetry Right now it's a bit of Mess
54:03
Well, what usually happens is gender releases is that as all most devs are running the rolling stuff It's not the most precious Yeah, I can't tell for sure about the instant media because last time I had to use the agent with some media Probably a very long time ago
54:21
Yeah, but I mean it still makes sense to tackle that at some point because I mean I'm an open while on an open wireless One I don't actually like to run emerge sync Any other finishing questions before we start I thought everybody installed from sisrescue CD anyway
54:47
I do for sure. Yeah
55:01
I haven't been able to boot it with EFI No, I haven't been able to Yeah, the comment the comment was that he uses Ubuntu because it boots better from EFI
55:25
Okay. Thanks, everyone