[procps] Re: rhbz#737215 - ps does not resolve some user names
- From: Jim Warner <james.warner@xxxxxxxxxxx>
- To: procps@xxxxxxxxxxxxx
- Date: Thu, 27 Sep 2012 15:08:15 -0500
On Sep 18, 2012, at 5:26 AM, Jaromir Capik <jcapik@xxxxxxxxxx> wrote:
> I'm unsure now, but I think that NTLM UIDs go from the upper limit
> down to zero (probably to avoid conflicts with local UIDs???).
> So ... even the very first user in the list can have
> a very long number. I'm a bit unsure if it has 24 bits or more.
Hi Jaromir,
I think I've solved the potential overflow problem, and a little bit more. But
rather than burden everyone with a rather large tarball, I've just sent the
patches to Craig.
Below is a copy of my covering letter. After Craig pushes the changes, I look
forward to any feedback.
Regards,
Jim
--------------------------------------------------------------
Begin forwarded message:
> From: Jim Warner <james.warner@xxxxxxxxxxx>
> Subject: ps/top truncation issues
> Date: September 27, 2012 2:56:23 PM CDT
> To: Craig Small <csmall@xxxxxxxxxx>
>
> Hi Craig,
>
> Thanks for applying my ps patch. Here's a brief summary of current ps/top
> truncation state:
>
> PS
> . now truncates USER type fields with '+'
> . never truncates PID/UID type numbers
> . but could overflow columns, misaligning others
> . that alerts user to need for explicit size override
>
> TOP
> . automatically adjusts many widths (PID, TGID, etc)
> . silently truncates USER types and other string fields
> . could overflow UID type numbers, misaligning others
> . there is no user remedy for such misalignment
>
> With the ps commit now applied, I consider him completed with respect to
> truncation. However, top's silent truncation and potential overflow both
> needed attention, so I tried to parallel ps. And while misaligned columns in
> ps are acceptable, under a continuously running program like top they would
> be a disaster.
>
> I have refactored top to eliminate any possible overflow/misalignment while
> providing the '+' visual clue when truncation occurs. If truncated, users
> can dynamically (temporarily?) widen affected fields if they wish. As I
> noted in a commit message, a user is free to vastly exceed screen dimensions
> with some undesirable but logical implications.
>
> Also, as a natural outgrowth of refactoring, column headings are now
> translatable and will be included in the .pot file. They had to be excluded
> previously because of their dependency on leading/trailing padding tied to
> the now non-existent format specifiers.
>
> Finally, I introduced separate justification capabilities for numeric and
> string columns. This feature may never be needed (or even used) but it
> required very little code, carried no additional runtime costs and was
> another benefit from refactoring (plus it was fun, see the postscript below).
>
> In spite of all the changes, performance remains on a par with versions 3.3.1
> through 3.3.3.
>
> The attached tarball contains all of the above plus 3 unrelated fixes.
>
> Regards,
> Jim
>
> p.s. It's been over a decade since any top feature earned a place of honor in
> the Section 7. STUPID TRICKS Sampler. With the new justification provisions,
> that long dry spell has finally been broken.
Other related posts:
