I'd like to make a snapshot soon, and asked Paul how his sort work was coming. He replied:
> 1 There's still a bug with sort -m -o f f; it can dump core > despite my recent patch there. I have a fix for this and > will push it (+ test case) shortly. > > 2 After further thinking about it, I now realize why there > used to be a reference count for process-IDs. It's possible > that two or more temp files are associated with the same process ID. > The old code was designed to detect this, but I don't think it worked. > The new code that I checked in a couple of days ago was not designed > to detect this (I considered the scenario, but incorrectly concluded > that it wouldn't lead to a problem), and I now know it doesn't work. > I have a fix in mind and will work on it next. > > 3 The sort --compress/hang problem. It's conceivable this is the > same bug as (2), but most likely it's different. > > None of these bugs are pressing: (1) has been there for ages > and (2) and (3) are present only if --compress is used, which > isn't common. So I don't think it matters much whether these > fixes are in the next version. I'd like (1) to go in at least, > and I think I can fix (2) in a day or two; dunno about (3). Thanks! I'll wait a little, in case a fix for (2) is forthcoming.
