Hi. just a few remarks to avoid confusion about the build system, we worked on.
please see inside the orcnotes below. rgds jan i On 24 November 2014 at 18:55, Dennis E. Hamilton <[email protected]> wrote: > Jan, > > Thanks for the very useful information. <orcnote>s below. > > > -- Dennis E. Hamilton > [email protected] +1-206-779-9430 > https://keybase.io/orcmid PGP F96E 89FF D456 628A > X.509 certs used and requested for signed e-mail > > -----Original Message----- > From: jan i [mailto:[email protected]] > Sent: Monday, November 24, 2014 03:24 > To: Dennis Hamilton > Cc: dev; [email protected]; jan iversen > Subject: Re: Capstone 2013 Client Requirements Document Status? > > On 24 November 2014 at 00:56, Dennis E. Hamilton <[email protected]> > wrote: > > > Oh my. I wasn't around when that project was created. Is there any > > update on what was accomplished? It ended before Summer 2014, yes? > [ ... ] > We ended having a pretty good converting tool for the simple function > groups, so in general this was more a proof of concept, than an actual > implementation. > > I learned enough to verify, that with a couple of students one more > semester we could have a microsoft visual studio solution running. > > <orcnote> > Is this work all done in Corvalis? That's not quite a day-trip for > Me, from Seattle, but I could visit while staying with my son in > Portland. > Yes as far as I know (steve knows the exact location). But work is finished, and we did not join this semester. Depending on my available time and interest on this list, we might join again next semester. It would be quite fun to finalize the ideas. > </orcnote> > > The project files dont need to be svn, we have tested to have Cmake > generator files, that can make both makefiles and project files. > > The whole subject was discussed earlier, and it was clearly stated that we > do not want an extra build system, on top of the unfinished ones we already > have, that is the main reason why the results never went further than the > wiki, the branch, and our google drive. > > <orcnote> > Great about keeping projects out of the SVN/Git repositories. I scanned > through other pages on the Build Project and learned a great deal on how > difficult it is to untangle the AOO structure at > <https://wiki.openoffice.org/wiki/Build_System_Analysis>. > The gyrations to set up a Windows compile are amazingly gnarly and it is > amazing how many little details have to be conquered, as shown by the > good work at > < > https://wiki.openoffice.org/wiki/Documentation/Building_Guide_AOO/Step_by_step#Windows_7 > > > and other pages, all having pieces of the story that need to be > consolidated somehow. > We actually worked on the idea of seperating build files and source code, meaning that you in parallel to our source structure would have a build directory. The build directory would consist of a fixed part (in svn/git) and a dynamic part, generated specifically for the platform you work on. > > The problem with CMake seems to be that it works for the complete build > but that doesn't allow individual construction and testing of modules. > There are some significant problems having dependencies such that a > commonly used module is not rebuilt every time a module that depends > on it is rebuilt. This is described as the Recursive Make problem and > Having a topographical ordering that does no redundant builds is sought. > Do I understand that correctly? > I did not express myself correctly. CMake would only be used to generate makefiles or project files for the platform you work on. A bit similar to the current configure. Currently our build system have fixed makefiles (in svn), and tons of environment variables to control the different platforms. Generating makefiles specifically for a platform, reduce build time with 10-15% (based on real life tests). By generating makefiles (or project files) in a seperate directory (incl. object files), you can have multiple platforms (e.g. debug, production) using the same sources. > > There is also the problem of version dependencies in whatever VS projects > and solutions CMake produces. > There are some problems here. I am a fan of CMake to describe what to generate, but admit that other people are in favour for a XML description file, that generate using a XSLT. Both ways have advantages and disadvantages, but both ways also secure a clear seperation of build items and source. > > I wonder if this is one of those problems for which one more level of > indirection might be quite useful. Just a thought. > A good thought....I think the expression is "two souls one thought" :-) > </orcnote> > > > > > > But I think clean builds using Microsoft tooling is a great idea, and it > > would be great if it did not require a POSIX shim and the friction that > > creates for what developers on the Windows platform are taught, know, and > > have free tools for. It could be all right to use a POSIX shim by those > > whose toolcraft favors it, but it should not be required. > > > I do too....goal was to get rid of cygwin, and work like a windows > programmer works. > > <orcnote> > I sympathize with having counterparts of POSIX/Gnu Tools and perhaps > still needing them for computational tasks that are part of the current > build process. But most of those have Windows native counterparts but > for command-line and file-system reference notation differences. One > Reason I want a half-step using MSYS2 was that it will run native Windows > tools just fine and that means one could also have Windows batch files > for such things as firing up the VC++ compile chain in a way that does > not require messing with the Windows registry, etc. > > "Working like a windows programmer works" may cast a net too wide. I > agree that there is an important on-ramp via Visual Studio and the > Windows tools. I do worry about the level of understanding of the > platform that is needed to approach something as complex as AOO. I tend > to want builds to be as simple as possible and having as few pre- > requisites as possible. I have been working to separate the edit, > compile, test, rinse repeat process, for which an IDE is ideal, from > the simpler build process by which another can replicate the build, > either because that is all they want to do or they are doing it to > support working on a different part of the source. > </orcnote> > > > > The availability of the Visual Studio 2013 Community Edition can have a > > great impact on this idea, since that now includes ATL, MFC, and much > more, > > including built-in git support. > > > yes, and please do not forget that all committers can get the full version > for free together with a bunch of other microsoft tools. > > <orcnote> > That committers have, for free, something that other individual > developers must pay for is not a solution to me. Likewise if only > students at recognized institutions get the tools for free. I think > the Community Edition may cure that, although it is pretty heavyweight > (9 GB and needs recent Windows version). It would be great if the > Visual Studio 2013 Express for Desktop were sufficient. I > suspect that, going forward, the Community Edition will be the > likely avenue for VS 2015 or certainly beyond 2015. The CE license > is also extremely friendly for use on open-source projects. > You have a very valid argument....of course I am just trying to get more committers :-) Actually we saw that the difference in the vcproj file between 2013 express, VS2015 and the community edition was something a XSLT dealt with easily. > </orcnote> > > > > > > A REQUEST: I'd like to see a module that was converted along the lines of > > this project to see if my ideas can also be applied to it. > > > you might want to check the capstone branch, some of the latest stuff > should be in there, otherwise I can privately share my google drive with > you. > > <orcnote> > GREAT! I am checking out the branch now. > </orcnote> > > Steve and I discussed continuing the project with this capstone semester, > but due to a lot of problems with their registration, and the somewhat > personal discussions on this and other AOO lists, I decided to take a > break. > > I still hope we can make a future project with capstone, because I at > least, find this to be a cornerstone in acquiring and keeping new > developers. Who knows in the future I might find others who share my point > of view. > > <orcnote> > I am completely aligned with regard to the cultivation of new > developers and providing a welcoming on-ramp for the learning curve > and toolcraft required. > > I had gone absent from the project and lists while you were > arriving, so I missed those discussions. Thank you for helping > me get up to speed about it. > My pleasure, that is why we are here, to make 1+1 == 3. > </orcnote> > > rgds > jan i. > > > [ ... ] > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [email protected] > For additional commands, e-mail: [email protected] > >
