Date: Tue, 9 Feb 2016 10:12:45 -0500 From: Jim Ohlstein <jim@ohlste.in> To: John Marino <freebsdml@marino.st>, Hrant Dadivanyan <hrant@dadivanyan.net> Cc: FreeBSD Ports ML <freebsd-ports@freebsd.org> Subject: Re: Removing documentation Message-ID: <56BA01ED.7000504@ohlste.in> In-Reply-To: <56B9F2D6.1090107@marino.st> References: <E1aT6jw-000MGn-1T@pandora.amnic.net> <56B9D609.6030407@marino.st> <56B9EDC7.1010403@ohlste.in> <56B9F2D6.1090107@marino.st>
next in thread | previous in thread | raw e-mail | index | archive | help
On 2/9/16 9:08 AM, John Marino wrote: > On 2/9/2016 2:46 PM, Jim Ohlstein wrote: >> After all of this "discussion" I decided to give synth a try. I have no >> pony in this race as I use neither portmaster nor portupgrade. Both may >> still be in my repo, but they are not installed. > > Thanks for trying it! > >> >> The build time of "like 20-30 minutes, at most" is ummm... let' just >> call it optimistic. I only needed five new dependencies. Poudriere was >> unable to take advantage of more than two parallel builders except for a >> rather short overlap where it used three, if I recall correctly. The >> vast majority of the time it used only one builder. Build and package >> time for gcc6-aux was 34:52 on an Intel E5-2650 v3. Build and package >> time for binutils, required for gcc6-aux, took 4:44. That's pretty close > > hmmm, my core i5 builds it in 10-12 minutes and I've had it ~4 years? > I'm not sure why such a big descreptancy, but newer machines with 4-8Gb > or more ram should have no issues with time. Interesting. I just tested again and got 31:47. To be fair, I only allow that VM four cores. Perhaps it might do better with more? > >> to 40 minutes for just two dependencies, one of which is a dependency of >> the other. Build and package time for synth was 1:09. >> >> I installed synth and had a look at the man page. Nice job on the >> documentation though I might suggest more real world examples, in an >> "Examples" section at the end, would be helpful to people like me who >> want to understand how to get started. Sort of like a "quick start >> guide" that comes with a new electronic component. Get it going and then >> read the details on what's really important for the specific use case. >> That shouldn't be construed as a knock on the documentation, which >> really is very good. > > Do you think the illustrated README on the github page is helpful? > > https://github.com/jrmarino/synth > Yes, I do indeed. Thanks. > >> This was last night and I haven't tried building with it yet. I need to >> re-read the documentation. I do however have concerns, the biggest of >> which is, yes, the dependencies. I use poudriere because I like to build >> packages myself for my installations and with my options, so using the >> FreeBSD repo version of synth will be a non-starter. That means that >> I'll need to rebuild gcc6-aux every time I need to rebuild synth, >> assuming gcc6-aux has been updated. It's a fair guess that gcc6-aux is >> regularly updated (the current version is dated 20160124). It's also a > > After gcc6 hits release (6.1), it will probably only be released with > every point release (6.2, 6.3), which are separated by months. > >> fair guess that synth will go through a few iterations in the short term >> given its youth. Looking at my recent build logs, the longest builds I > > It's feature complete and v1.00 is coming out in a few days (same as > 0.99_6 with a version bump). v1.1 will come soon after when I improve > on the build-repository command to not scan the entire tree. It's > already been through the iterations, so I don't there will be that many > more. In any case, it's a small problem (and when gcc6-aux is released, > it won't build the bundled libraries anymore but use other ports so it > will be much, much faster to compile. > > >> run are far shorter than 35 minutes. This will slow things down and I'm >> not certain I'm going to be willing to keep a package in my repo that >> requires that amount of build time just as a dependency I otherwise >> would never build. To be honest, synth, which I will try, will have to >> be _far_ superior to poudriere in order to replace it as my tool of >> choice. Of course that's my use case and mine only. > > To be fair, poudriere users aren't the target audience. Yes, it's > significantly faster than poudriere and maybe people like the interface > better, but if they are already set up on poudriere and happy with it, > that's a fine choice too. > > It's more for people that aren't using poudriere, really should be, but > are intimidated by it. > -- Jim Ohlstein "Never argue with a fool, onlookers may not be able to tell the difference." - Mark Twain
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?56BA01ED.7000504>