Ok, after fighting with svn for a while I finally figured out how to extract the actual patches pending for 0.9.30.1, and have attached a tarball containing each one so nobody else has to fight with the horrible svn branching. Each file in the tarball is "svn-#####.patch" starting with the patch description from svn and ending with the actual patch. They apply with -p0.
Here's a few low-hanging fruit, review-wise: svn 24023: small obscure build bugfix, ok. svn 24085: improve error reporting, ok. (The comment about using a linux- specific hack when checking to see if we correctly found the linux kernel headers seems kind of insane. Yes, the Linux kernel headers are from Linux. Well spotted. But that's a minor quibble...) svn 24089: mips assembly bugfix of some kind (32 vs 64 bit code), and removal of an unused inline function. Ok. svn 24095: move dl_iterate_phdr to where glibc puts it, as various package builds expect to find it there. No argument with that goal, although the implementation seems overcomplicated. (The file libc/misc/elf/dl-core.c looks amazingly unnecessary and adding two new makefiles to move a symbol seems kind of overkill. Oh well.) svn 24139: thumb2 pie bugfix. Ok. svn 24141: Fix two memset bugs in the locale generation plumbing. Ok. svn 24142: Typo fix in wchar support. Ok. svn 24232: More wchar/locale fixes. Ok. Presumably the compiler is smart enough to turn (pwcs[i] != (pwcs[i] & 0x7f)) into (wpcs[i] & (~0<<7)), or perhaps even pwcs[i] < 128 if wchar_t is unsigned... svn 24278: mips bugfix renaming a function from __NTH to __cmsg_nxthdr in bits/socket.h. Ok. No idea what this fixes, and "merge from trunk" isn't helpful, but it's small and localized and svn's horrible branch infrastructure doesn't give me enough info to argue _against_ it, either. Assuming whatever problem it addresses exists, this looks like a quite reasonable solution. svn 24326: add missing curly brackets to a for() loop, fixing getaddrinfo. Ok. svn 24589 and svn 24798: Make uClibc against a 2.6.28 kernel headers (first one does i386, second does x86_64). The most important bug fixes in the lot. These two by _themselves_ would demand a uClibc dot release. Yes please. svn 24836: Instead of conditionally exporting PIE flags in Rules.mak, make their _use_ conditional in utils/Makefile.in. Dunno what bug this fixes, but it's cleaner. Ok. svn 24839: Use -pie instead of -Wl,-pie so gcc uses the right start/end code, and turn __dl_ldso__ into a hidden weak symbol for some reason. Ok. (It says I should fire up svn again and look at the description of some totally unrelated commit number in a different branch, and I'm not gonna.) svn 24945: Typo fix in #ifdef header guard. Ok. (Why didn't this break more obviously earlier?) svn 24949: Typos in comments. Of vital importance to get this in the bugfix- only dot release, I'm sure. (On the bright side, just about mathematically guaranteed to NOT MATTER to the resulting binary, so no reason to object other than that it's noise.) svn 24956: Description says move linker flags around to work around a gentoo bug, and I'll believe it. Ok. svn 24963: Remove erroneous hidden_def()s. Ok. Aside: Why the heck does libm have its own libm_hidden_def() macro? Is there any reason at all not to just use hidden_def() directly, since the header goes #define libm_hidden_def() hidden_def()? Should we have a #define libm_printf() printf() and use that there too? What the...? svn 25229: cross compiling fix so install_headers doesn't freak about unsupported linker options (since you aren't linking when you install the headers). That leaves svn 24052 and 24489, which I'll deal with individually in later emails. (Those two may not qualify as bugfixes.) Rob
patchlist-0.9.30.1.tar.bz2
Description: application/bzip-compressed-tar
_______________________________________________ uClibc mailing list [email protected] http://lists.busybox.net/mailman/listinfo/uclibc
