posix threading plans

[Joseph S. Myers]

On Sun, 6 May 2007, Mike Frysinger wrote:

> the plan for NPTL is to get the extraneous ports and patchsets (like arm and 
> sh) floating around merged into the uClibc-nptl branch.  then we need to get 

I'm still concerned about the incomplete and undocumented status of merges 
from trunk into this branch.  If I diff libc/sysdeps/linux/arm/, say, 
between trunk and branch, I see several changes that are reversions of 
patches made on trunk months ago, despite the last "Merge from trunk" 
commit being on 2007-04-21.  This is the issue I explained at 
<http://www.uclibc.org/lists/uclibc/2006-August/016155.html>: unless every 
single merge commit (a) merges every change in the relevant range of trunk 
revisions and (b) states in the log message what those revisions are, it 
becomes pretty much impossible afterwards to work out what isn't merged, 
and so which diffs are real changes on the branch (and should be merged to 
trunk) and which are reversions of changes not merged from trunk (and such 
reversions mustn't be merged to trunk).

For this reason, I still think a separate branch based on current trunk, 
getting only individual logical patches that have be checked line-by-line 
applied to it, and with no partial merges from trunk made to it, is the 
way to go.

The ARM NPTL work was based on trunk at that time (and subsequently merged 
from newer trunk) precisely because the incomplete and undocumented merges 
made it infeasible to work based on the NPTL branch without getting 
regressions relative to trunk.

-- 
Joseph S. Myers
joseph at codesourcery.com