I'm going to merge several discussions from several places to here - I hope you don't mind toad.
Matthew Toseland skrev: > I am not in a position to maintain it without help - I have not tried to > build > it yet. It uses one unmaintained freeware utility in the build process and > another one at install time. And it is not localisable, and has to force font > sizes and so on, with the result being that it likely would require major > changes for l10n. In short it's not ready. Thanks for your help, but I'm not > in a position to fix it and make it ready at the moment. IMHO it is not a > critical piece of functionality for 0.8.0. What *is* critical is that > start/stop work on Vista... getting clean start.exe/stop.exe's that can be > shipped with our current installer is not straightforward either, but it does > avoid some of the other problems... > Matthew Toseland skrev: > Using unmaintained freeware in the build process is unacceptable. > > Source is available for XN Resource Editor, is there a command line version? > > http://www.wilsonc.demon.co.uk/d10resourceeditor.htm > > Also, RemProf.exe is freeware and not maintained and source code will not be > made available, and we use that at install time. > > Do we use any freeware utilities in the current java installer? I think maybe > there was one...? If so, maybe it's an acceptable risk, but it should be > minimised nonetheless - what we need most urgently are UAC-escalated > start/stop exe's for the java installer. Matthew Toseland skrev: > In fact, we already use netuser.exe on Windows ... see bug #1709. This is > closed source freeware, although we may be able to replace it one day as it > doesn't do very much; you probably call the dll's in your AHK code. > > Anyway, mingwin comes with a resource script compiler. So all we need to do > is > extract the resource script (which is source code), and if necessary edit it, > then compile it with mingwin and feed it to the AHK compiler? This eliminates > the need for reshacker.exe, so we can build start.exe and stop.exe cleanly > from source and use them in our installer. Can AHK compile resource scripts, > or does it need it to be already compiled? Could you extract and decompile, > or assemble, the resource file needed for start.exe/stop.exe to escalate > themselves? > > If we can fix the fact that Vista doesn't work before 0.8.0, that would be > great. But actually using the AHK-based installer has other issues, so *for > now* I'm saying no. > I plan on maintaining the installer regardless of my real life plans. But I'm not sure that I will have to time to implement major features during my military service (depends on how much free time I get in the next 4 months that it lasts). Real life comes first, after all. Regarding third party utilities, I think you need to look at the situation from another perspective. Here is my viewpoint: What we currently use in the Java installer: - cat.exe (used by install scripts, open-source GNU) - netuser.exe (used to set "password never expires" flag on an account. Unknown origin (credited "Siemes AG") and source status) - Ntrights.exe (used to set account rights. Seems to come from a Windows 2000 resource kit (credited "Beta Version by Georg Zanzen"). Probably unmaintained and closed-source?) - sed.exe (used by install scripts - wget.exe (used by install scripts, open-source GNU) In my installer: - cat.exe, sed.exe and wget.exe are not needed - netuser.exe and Ntrights.exe usage has been copied/ported So far, that must be kind of an improvement in this area, aye? When testing, I realized that the user's profile folder ("C:\Documents and Settings\..." on XP) and various user registry entries are *not* deleted when we delete the user in the current installer, and that there apparently isn't a straightforward way to do this (yes, another argument from me about not creating a custom user account - but never mind). Instead of leaving the files intact (cluttering the user's machine + leaving permanent traces of Freenet), I found a third party freeware utility called RemProf.exe from an IT consultancy website. In my opinion, it's more important to have full functionality with the cost of using a couple of closed-source tools rather than not providing full functionality, and in this case, leaving permanent and easily-available traces of Freenet on every single Windows machine it has been installed onto. If you don't agree - I can simply remove the utility again. In any way, I cannot see how this can be an argument *against* deploying my installer, as worst-case scenario (when comparing to the current installer) is 'no improvement' in this particular area. In the build process, I'm using the closed-source freeware tool Resource Hacker (http://www.angusj.com/resourcehacker/) to add Freenet icons and Vista UAC manifests to the binaries. Resource Hacker is pretty much *the* tool to modify Windows executable resources. If you don't take my word on it, googling "Resource Hacker" returns ~775.000 results (you could compare that to the ~19.000 results for "ntrights"). The 3 million downloads (according to the site) kind of supports that. It is true that it is no longer under any development, but I'd argue that it doesn't matter, since it's practically "done" and works just fine. There are most likely open-source tools with the same functionality (you mention a MinGW tool yourself), but in my opinion it was a better solution to use the well-known tool I knew worked and knew how to use, compared to spend my limited time testing new tools and delaying a release. Regarding l10n, I don't know where you've got: "...And it is not localisable, and has to force font sizes and so on, with the result being that it likely would require major changes for l10n...." from. I have chosen to set a specific font and font size for the english version to ensure GUI layout consistency across different setups - rather than relying on unknown user settings that might mess up the layout. A translator can simply pick another font and font size for his language if he feels that works better. Of course implementing l10n requires changes to the source (as any feature does). Again, I've felt it was more important to actually get a good, working installer out, rather than spending a lot of time in l10n which always can be added later on. As far as I know, the greatest commonly understood language among our users is English, hence it makes sense to me to start out with an English version. Regarding "Can AHK compile resource scripts, or does it need it to be already compiled?": The AHK compiler is shipped with a "base" consisting of the library and PE headers and stuff - and then compiles the script into/on top of the base. The base is what we modify to replace and add resources (Freenet icons and UAC manifest). Regarding "Could you extract and decompile, or assemble, the resource file needed for start.exe/stop.exe to escalate themselves?": The manifest that needs to be "manifested" into an .exe to make it selv-UAC-elevate is in SVN already. In my very humble opinion, we are talking about *minor* issues compared to great improvements. But I'll leave that to you guys to decide... - Zero3 _______________________________________________ Devl mailing list [email protected] http://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl
