Sort all items in the tarball stats report

[webbnh]

Currently, the v0.69 branch testing is failing due to a non-determinism in the server/bin/pbench-tarball-stats.py output. For parts of the report, the lists of servers are sorted; and, for other parts they are not. This change adds the missing sort directives to the output template.

Fixes #2623.

[dbutenhof]

My first reaction is, "oh, no, another revlock" ... however, realistically, we need real configuration management which means everything is revlocked to not just a range but a specific version, preferably from a supported RPM. So, yeah. Cool. Third time's a charm?

[dbutenhof]

Oops: Or,

ERROR: Invalid requirement: 'humanize>=3.11 <4.0.0' (from line 3 of /home/pbench/server/requirements.txt)

[webbnh]

"oh, no, another revlock"

Yeah, well, adding the rev-lock actually has nothing to do with the problem I'm fixing...it's just that when I went to test my change, it suddenly failed in a new way...because, apparently, the build picked up a new version of the humanize package which had recently rolled from v3.11 to v4.0 and put in a breaking change. And, of course, the breakage isn't in our code -- it's in jinja2-humanize-extension which we depend upon -- but we inherit the problem. (I reported the issue, but I'm dubious as to whether it will be fixed -- jinja2-humanize-extension seems to live in a very quiet repo.) So, I added the rev-lock (and, as you noticed, it took two tries...).

And this enabled me to see that the resulting sort comes out with the opposite order from what I expected, so I have yet another update with the fixed gold file. 😛 But, it looks good, now (modulo what LGTM comes up with...).

[thebaptiste]

The repo jinja2-humanize-extension is not so quiet...
We have fixed the issue requiring humanize<4.0.0 in release 0.2.1
We will try to phase with humanize 4.0.0 later
Thanks for the issue

[webbnh]

Hi @thebaptiste! Thanks for commenting here!!

I apologize for my earlier comment -- clearly my doubts were misplaced: the fact that my issue had a single-digit number must be an indication of the quality of the work on jinja2-humanize-extension rather than any lack of attentiveness from its maintainers or users. 😀

(I'm quite amused by the fact that my referencing it in this repo added a mention on the Issue in your repo...it didn't occur to me that GitHub provides that level of integration. 😄)