Re: Qt packaging
On Fri, Jul 22, 2022 at 11:03:54AM +0200, Jan Rękorajski wrote: > Can someone explain why are we using split sources/packages for Qt? Beside build time and space requirements, I see one more reason now: rebuilding after dependent package soname change is more painful. Current case: jasper 3.x. Split case: just qt5-qtimageformats.spec to rebuild Monolithic case: whole qt6.spec to rebuild -- Jakub Boguszhttp://qboosh.pl/ ___ pld-devel-en mailing list pld-devel-en@lists.pld-linux.org http://lists.pld-linux.org/mailman/listinfo/pld-devel-en
Re: Qt packaging
On 08.08.2022 23:34, Jan Rękorajski wrote: > On Mon, 08 Aug 2022, Jan Palus wrote: > > > On 08.08.2022 08:32, Jan Rękorajski wrote: > > > On Fri, 22 Jul 2022, Jan Palus wrote: > > > > > > > On 22.07.2022 11:03, Jan Rękorajski wrote: > > > > > Can someone explain why are we using split sources/packages for Qt? > > > > > > > > > > I want to add Qt6 and building from the monolythic source is s > > > > > much > > > > > easier. No need for bootstrap, no intertwined build dependencies, just > > > > > configure -> build -> build docs -> install. > > > > > > > > > > And unless there is a _very_ good reason to use split sources I'm > > > > > just going > > > > > to add a single qt6 package that builds everything (we can still > > > > > subpackage > > > > > bineries as we want them). > > > > > > > > As long as each component is bcondized and there are no "to the exact > > > > release" dependencies then I guess it's fine. Doing qtwebengine (and all > > > > the other components) rebuild each time qtbase needs a small packaging > > > > adjustment would be tough on arm, though I'd understand if nobody cared > > > > about my use case. > > > > > > FYI build time on builders is 1.5 hour without qtwebengine and 7 hours > > > with qtwebengine. > > > > > > I don't know how it looks on arm, but IMHO no-webengine bcond should be > > > enough? > > > > Multiply it by ~4 and it's roughly result for arm. The first part I > > mean, qtwebengine is so heavy that I build it in AWS. > > > > Anyway no worries, if needed I can add more bconds myself. And thanks a > > lot for working on qt6! > > Thanks, it's a slow and painful process, and we'll end up with less > granularity, at least at the beginning. What I want now, is a MVP to be able > to build current calibre :/ > > Out of curiosity, would webengine even build on arm? What I see it > building is full blown blink/chromium engine. And this thing has lots > of fancy dependencies, both software and hardware. Just to be clear when I said "arm" I really meant "aarch64", 32-bit version of qtwebengine is of not much interest to me. And at least qtwebengine 5.x builds just fine for aarch64 and using it daily as my primary web browsing engine. I don't expect qt6 to be much different but haven't tried to build it so far. ___ pld-devel-en mailing list pld-devel-en@lists.pld-linux.org http://lists.pld-linux.org/mailman/listinfo/pld-devel-en
Re: Qt packaging
On Mon, 08 Aug 2022, Jan Palus wrote: > On 08.08.2022 08:32, Jan Rękorajski wrote: > > On Fri, 22 Jul 2022, Jan Palus wrote: > > > > > On 22.07.2022 11:03, Jan Rękorajski wrote: > > > > Can someone explain why are we using split sources/packages for Qt? > > > > > > > > I want to add Qt6 and building from the monolythic source is s much > > > > easier. No need for bootstrap, no intertwined build dependencies, just > > > > configure -> build -> build docs -> install. > > > > > > > > And unless there is a _very_ good reason to use split sources I'm just > > > > going > > > > to add a single qt6 package that builds everything (we can still > > > > subpackage > > > > bineries as we want them). > > > > > > As long as each component is bcondized and there are no "to the exact > > > release" dependencies then I guess it's fine. Doing qtwebengine (and all > > > the other components) rebuild each time qtbase needs a small packaging > > > adjustment would be tough on arm, though I'd understand if nobody cared > > > about my use case. > > > > FYI build time on builders is 1.5 hour without qtwebengine and 7 hours > > with qtwebengine. > > > > I don't know how it looks on arm, but IMHO no-webengine bcond should be > > enough? > > Multiply it by ~4 and it's roughly result for arm. The first part I > mean, qtwebengine is so heavy that I build it in AWS. > > Anyway no worries, if needed I can add more bconds myself. And thanks a > lot for working on qt6! Thanks, it's a slow and painful process, and we'll end up with less granularity, at least at the beginning. What I want now, is a MVP to be able to build current calibre :/ Out of curiosity, would webengine even build on arm? What I see it building is full blown blink/chromium engine. And this thing has lots of fancy dependencies, both software and hardware. -- Jan Rękorajski| PLD/Linux SysAdm | bagginspld-linux.org | http://www.pld-linux.org/ ___ pld-devel-en mailing list pld-devel-en@lists.pld-linux.org http://lists.pld-linux.org/mailman/listinfo/pld-devel-en
Re: Qt packaging
On 08.08.2022 08:32, Jan Rękorajski wrote: > On Fri, 22 Jul 2022, Jan Palus wrote: > > > On 22.07.2022 11:03, Jan Rękorajski wrote: > > > Can someone explain why are we using split sources/packages for Qt? > > > > > > I want to add Qt6 and building from the monolythic source is s much > > > easier. No need for bootstrap, no intertwined build dependencies, just > > > configure -> build -> build docs -> install. > > > > > > And unless there is a _very_ good reason to use split sources I'm just > > > going > > > to add a single qt6 package that builds everything (we can still > > > subpackage > > > bineries as we want them). > > > > As long as each component is bcondized and there are no "to the exact > > release" dependencies then I guess it's fine. Doing qtwebengine (and all > > the other components) rebuild each time qtbase needs a small packaging > > adjustment would be tough on arm, though I'd understand if nobody cared > > about my use case. > > FYI build time on builders is 1.5 hour without qtwebengine and 7 hours > with qtwebengine. > > I don't know how it looks on arm, but IMHO no-webengine bcond should be > enough? Multiply it by ~4 and it's roughly result for arm. The first part I mean, qtwebengine is so heavy that I build it in AWS. Anyway no worries, if needed I can add more bconds myself. And thanks a lot for working on qt6! ___ pld-devel-en mailing list pld-devel-en@lists.pld-linux.org http://lists.pld-linux.org/mailman/listinfo/pld-devel-en
Re: Qt packaging
On Mon, 08 Aug 2022, Neal Gompa wrote: > On Mon, Aug 8, 2022 at 2:32 AM Jan Rękorajski wrote: > > > > On Fri, 22 Jul 2022, Jan Palus wrote: > > > > > On 22.07.2022 11:03, Jan Rękorajski wrote: > > > > Can someone explain why are we using split sources/packages for Qt? > > > > > > > > I want to add Qt6 and building from the monolythic source is s much > > > > easier. No need for bootstrap, no intertwined build dependencies, just > > > > configure -> build -> build docs -> install. > > > > > > > > And unless there is a _very_ good reason to use split sources I'm just > > > > going > > > > to add a single qt6 package that builds everything (we can still > > > > subpackage > > > > bineries as we want them). > > > > > > As long as each component is bcondized and there are no "to the exact > > > release" dependencies then I guess it's fine. Doing qtwebengine (and all > > > the other components) rebuild each time qtbase needs a small packaging > > > adjustment would be tough on arm, though I'd understand if nobody cared > > > about my use case. > > > > FYI build time on builders is 1.5 hour without qtwebengine and 7 hours > > with qtwebengine. > > > > I don't know how it looks on arm, but IMHO no-webengine bcond should be > > enough? > > > > The reason most distros don't use the monolithic source is that it's a > pain to apply patches to it. Qt doesn't actually get developed that > way, and backporting fixes is more of a pain if you use the monolithic > build. Well, we don't have resources to play with backporting changes. Besides I saw have ex. Fedora packages qt and it is IMO a joke. They don't build docs, they don't build internal deps, so yeah, it's easy, but it's half of the functionality. I'd rather have a package without the backports, but with all bells and whistles, that is easy to build, rather than either build pain on half baked. -- Jan Rękorajski| PLD/Linux SysAdm | bagginspld-linux.org | http://www.pld-linux.org/ ___ pld-devel-en mailing list pld-devel-en@lists.pld-linux.org http://lists.pld-linux.org/mailman/listinfo/pld-devel-en
Re: Qt packaging
On Mon, Aug 8, 2022 at 2:32 AM Jan Rękorajski wrote: > > On Fri, 22 Jul 2022, Jan Palus wrote: > > > On 22.07.2022 11:03, Jan Rękorajski wrote: > > > Can someone explain why are we using split sources/packages for Qt? > > > > > > I want to add Qt6 and building from the monolythic source is s much > > > easier. No need for bootstrap, no intertwined build dependencies, just > > > configure -> build -> build docs -> install. > > > > > > And unless there is a _very_ good reason to use split sources I'm just > > > going > > > to add a single qt6 package that builds everything (we can still > > > subpackage > > > bineries as we want them). > > > > As long as each component is bcondized and there are no "to the exact > > release" dependencies then I guess it's fine. Doing qtwebengine (and all > > the other components) rebuild each time qtbase needs a small packaging > > adjustment would be tough on arm, though I'd understand if nobody cared > > about my use case. > > FYI build time on builders is 1.5 hour without qtwebengine and 7 hours > with qtwebengine. > > I don't know how it looks on arm, but IMHO no-webengine bcond should be > enough? > The reason most distros don't use the monolithic source is that it's a pain to apply patches to it. Qt doesn't actually get developed that way, and backporting fixes is more of a pain if you use the monolithic build. -- 真実はいつも一つ!/ Always, there's only one truth! ___ pld-devel-en mailing list pld-devel-en@lists.pld-linux.org http://lists.pld-linux.org/mailman/listinfo/pld-devel-en
Re: Qt packaging
On Fri, 22 Jul 2022, Jan Palus wrote: > On 22.07.2022 11:03, Jan Rękorajski wrote: > > Can someone explain why are we using split sources/packages for Qt? > > > > I want to add Qt6 and building from the monolythic source is s much > > easier. No need for bootstrap, no intertwined build dependencies, just > > configure -> build -> build docs -> install. > > > > And unless there is a _very_ good reason to use split sources I'm just going > > to add a single qt6 package that builds everything (we can still subpackage > > bineries as we want them). > > As long as each component is bcondized and there are no "to the exact > release" dependencies then I guess it's fine. Doing qtwebengine (and all > the other components) rebuild each time qtbase needs a small packaging > adjustment would be tough on arm, though I'd understand if nobody cared > about my use case. FYI build time on builders is 1.5 hour without qtwebengine and 7 hours with qtwebengine. I don't know how it looks on arm, but IMHO no-webengine bcond should be enough? -- Jan Rękorajski| PLD/Linux SysAdm | bagginspld-linux.org | http://www.pld-linux.org/ ___ pld-devel-en mailing list pld-devel-en@lists.pld-linux.org http://lists.pld-linux.org/mailman/listinfo/pld-devel-en
Re: Qt packaging
On 22.07.2022 11:03, Jan Rękorajski wrote: > Can someone explain why are we using split sources/packages for Qt? > > I want to add Qt6 and building from the monolythic source is s much > easier. No need for bootstrap, no intertwined build dependencies, just > configure -> build -> build docs -> install. > > And unless there is a _very_ good reason to use split sources I'm just going > to add a single qt6 package that builds everything (we can still subpackage > bineries as we want them). As long as each component is bcondized and there are no "to the exact release" dependencies then I guess it's fine. Doing qtwebengine (and all the other components) rebuild each time qtbase needs a small packaging adjustment would be tough on arm, though I'd understand if nobody cared about my use case. ___ pld-devel-en mailing list pld-devel-en@lists.pld-linux.org http://lists.pld-linux.org/mailman/listinfo/pld-devel-en