| derRichard | stekern: does clang not support it or does it need some openrisc related work? | 00:06 |
|---|---|---|
| derRichard | transparent_union is not really a new thing. but it's rather unknown (it was also to me) | 00:06 |
| stekern | derRichard: clang supports it. but does it wrong (casts the function argument to union and threat it as such instead of using the calling convention of the first member of the union) | 06:12 |
| stekern | it's not openrisc specific | 06:13 |
| derRichard | stekern: ufff. have you filed a bug? | 10:26 |
| stekern | not yet, I'll probably ask on the cfe-dev list soon | 10:49 |
| stekern | it probably works ok on x86 since it expands small records | 10:50 |
| derRichard | maybe | 10:51 |
| stekern | one thing that is a bit unclear in the gcc description of the transparent_union is what should happen if you pass the union itself as an argument | 11:00 |
| jeremybennett | stekern: Good question for amylaar when he's on. He's a bit of a C language lawer. | 11:01 |
| stekern | yeah, I guess it's more of a hypothetical question, it's not intended to be used that way | 11:24 |
| derRichard | as transparent_union is gcc specific a c language lawer won't help much :) | 11:26 |
| mbuf | is there a gcc toolchain (that uses glibc) available for OpenRISC? | 12:12 |
| jeremybennett | No port of glibc yet. There are uClibc and newlib implementations. | 12:14 |
| derRichard | mbuf: the gcc port does not support dynamic libs. so no glibc | 12:15 |
| mbuf | derRichard: jeremybennett I see | 12:16 |
| mbuf | I just joined the openrisc mailing list | 12:16 |
| jeremybennett | derRichard: you can construct a static linked glibc | 12:16 |
| jeremybennett | support for dynamic linking is a separate issue - it also affects uClibc. | 12:16 |
| mbuf | jeremybennett: I see | 12:17 |
| jeremybennett | mbuf: What do you want glibc specifically? | 12:17 |
| derRichard | jeremybennett: AFAIK you cannot build glibc without a dyn linker | 12:17 |
| mbuf | jeremybennett: are there any TODO tasks related to toolchain that one can start to help with? | 12:17 |
| mbuf | jeremybennett: if one wants to build bash and other RPMs | 12:17 |
| jeremybennett | derRichard: Do you mean "build" or "use". The entire OpenRISC tool chain is run on a separate host (it is a cross compiler). | 12:18 |
| derRichard | jeremybennett: build and use :P | 12:18 |
| mbuf | BRB | 12:20 |
| jeremybennett | You can certainly build a static glibc. The only question is whether building glibc-static requires a dynamic glibc be built first. | 12:23 |
| derRichard | jeremybennett: AFAIK it does | 12:23 |
| jeremybennett | Looking at what others have done, it seems you need a few extra tweaks beyond just --disable-shared in the configuration, to deal with various utilities. | 12:23 |
| derRichard | glibc is everything else than embedded friendly :D | 12:23 |
| jeremybennett | derRichard: agreed - hence my question of why mbuf wanted glibc! | 12:24 |
| derRichard | if you want "real" userspace you have to have glibc | 12:24 |
| jeremybennett | What do you mean by "real" | 12:25 |
| derRichard | e.g. a sane libpthread | 12:25 |
| jeremybennett | Hmmm.... | 12:25 |
| derRichard | sane futex() support... | 12:25 |
| jeremybennett | I think that glibc has a better implemenation of libpthread than uClibc, but I am not sure there is a causal dependency. | 12:25 |
| derRichard | *much* better | 12:26 |
| jeremybennett | Similarly with futex. | 12:26 |
| jeremybennett | Basically by being much bigger glibc doesn't have to cut corners. But I think if you really wanted just good pthread or futex, you could achieve that in an embedded library by compromising elsewhere. | 12:27 |
| jeremybennett | As far as I can see uClibc has not bee very actively developed for some years. There are alternatives around, but I don't have a good feel for whether any of them are better. | 12:28 |
Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!