| |||||||||||||
|
| |||||||||||||
64 bit?64 bit?Posted Aug 30, 2005 21:09 UTC (Tue) by evgeny (guest, #774)In reply to: 64 bit? by felixfix Parent article: The second OpenOffice.org 2.0 beta
The fact that after three or so years, OO still doesn't event compile (let alone run...) under a 64-bit OS, says a lot about the quality of the code... From my experience, "porting" to 64-bits (which usually means fixing bugs made implicitly assuming sizeof(int) == sizeof(void *) etc) doesn't take too much time. So what a mess the OO codebase must be... shrugging.
(Log in to post comments)
64 bit? Posted Aug 31, 2005 3:13 UTC (Wed) by proski (subscriber, #104) [Link] I think it's a good style to check facts before belittling someone in public. OOo builds and runs, it's just not very stable. See e.g. http://artax.karlin.mff.cuni.cz/~kendy/blog/. Sure, if you search for "openoffice x86_64" you'll get many problem reports because users try to build OOo when they can download the binaries. You can notice that most reported problems are generic build problems (e.g. linker using wrong libraries), rather than compile errors.Compile errors is the easy part, and it has been done. Debugging and fixing problems that don't result in compile errors is the hard part, and it's going on now. Unfortunately, this effort is not helped by the fact that OOo is somewhat closed to occasional external contributors. If only they used a widely accepted version control system instead of CWS! On the lighter note, 64-bit support seems correllated with version control systems. When the repository is widely open to the public, everybody comes with their favorite kinds of patches. Somebody fixed 64-bit support after fixing it in 64 other projects, somebody polishes the GUI after doing the same in Gnumeric, somebody checks the code with Valgrind. I'm yet to see a project hosted on Subversion that would not compile cleanly on AMD64 :-)
64 bit? Posted Aug 31, 2005 10:32 UTC (Wed) by evgeny (guest, #774) [Link] > I think it's a good style to check facts before belittling someone in public.
Agreed.
> OOo builds and runs, it's just not very stable.
So I assume, by your strict adherence to facts, you actually built it yourself (under a 64bit OS); and you use it daily but it's just a bit less stable than the i386 build, right? Then you must be a lucky person (or maybe it's just me who is out of luck completely).
> See e.g. http://artax.karlin.mff.cuni.cz/~kendy/blog/.
So, someone periodically manages to compile the damned thing with half of the features turned off which then survives a few keystrokes until it painfully dies; and even the fixes needed for this pathetic advance are periodically rejected/overriden/whatnot upstream. Is that what you call "just not very stable"?
> Sure, if you search for "openoffice x86_64" you'll get many problem reports because users try to build OOo when they can download the binaries.
Which binaries? Please show me where I can download an x86_64 OO build.
> You can notice that most reported problems are generic build problems (e.g. linker using wrong libraries), rather than compile errors.
Yes, because the configuration/build system is completely screwed up. It requires a dozen of tools (perl/java/ant/jython/tcsh/...) installed only to advance to the stage when you can actually start compiling the monster. And then, after setting a hunderd of envirement variables something (e.g. ant) stops working (which runs fine otherwise). No wonder rare brave knights come past this quest.
> Debugging and fixing problems that don't result in compile errors is the hard part, and it's going on now.
It's been going for years; and apparently, 2.0 will still be 32-bits-only.
> If only they used a widely accepted version control system instead of CWS!
I think this is a pseudo-correlation (or whatever the correct mathematical term is), like the fact that people using skin sunscreens die of skin cancer more frequently. Which is not the cancerogenic nature of the sunscreens, of course (though there have been yellow press pubs hinting at this), but a one-more-level-deep true correlation (people that like to tan a lot often use sunscreens, which don't shield 100% of the UV rays, so the chances to get cancer are still higher than for those who don't tan). Same is here. True, in spirit, FLOSS projects tend to use convenient VCS'es. And, in a true FLOSS project, any 32bit archaisms get noticed and fixed at the early stages of development, before they plague the codebase in thousands of places.
The fact is that among open-source porjects of at least minimal public exposure, OO remains the only one which is not portable yet.
|
|
||||||||||||