New upstream version 3.3.36-1.upstream/3.3.36-1

1 files changed, 196 insertions, 0 deletions

diff --git a/inxi.changelog b/inxi.changelog
index 51a5202..9022a79 100644
--- a/inxi.changelog
+++ b/inxi.changelog
@@ -1,4 +1,200 @@
 ================================================================================
+Version: 3.3.36
+Patch: 00
+Date: 2024-09-04
+--------------------------------------------------------------------------------
+RELEASE NOTES:
+--------------------------------------------------------------------------------
+
+Phase 2 of the big CPU upgrade a few years back is now done, I'd left one part 
+inadequate in terms of the data structures, mainly because I did not have data 
+samples to test, but also because no corner cases that required a more robust 
+data structure for die > cluster > core counts showed up duing the initial 
+development phase. This led to somewhat predictable issues and bug reports when 
+someone had a CPU that did require that structure to show correct core/die type 
+data.
+
+So while most users will see very little difference, beyond dies: and clusters:
+items appearing where they did not before, the internal logic is now far more 
+robust.
+
+--------------------------------------------------------------------------------
+SPECIAL THANKS:
+
+1. CPU: Codeberg issue #307 finally helped solve the old AMD ryzen wrong core 
+count issue, I believe. Thanks superkoning for the data and reporting the 
+RISC-V failure to show right core counts, and to show MT when it isn't, which
+was caused by same issue.
+
+2. RECOMMENDS: Codeberg user Ricky-Tigg in issue #309 for taking the time to 
+actually check --recommends output for inconsistencies. And finding them!
+
+--------------------------------------------------------------------------------
+KNOWN ISSUES:
+
+1. The old issue #293 about AMD Ryzen 2 die CPU showing wrong core counts is 
+probably fixed, but can't know until it's confirmed for that exact CPU. Fix 1a 
+should handle all variants, I hope. Tested on old OS VM installs, and found most 
+of this logic never fires since /sys had a very barebones topology section f for 
+CPUs back then.
+
+2. DESKTOP: No known way to get Cosmic DE version, or Iced toolkit version. If 
+you know of a non-hackish way to get that version info, let me know.
+
+3. DRIVES: The disk vendor pinxi/tools/lists/disks.unhandled list keeps getting 
+longer, but most are not possible match, either not unique, or unknown models.
+Doing these matches is incredibly tedious, so I only do it every few releases, 
+but this time there were a LOT of new matches and vendors.
+
+--------------------------------------------------------------------------------
+BUGS:
+
+1. CPU: wrong core counts due to inadequate internal logic, and some new uses of
+clusters to categorize core ids. Since this is a pretty new change in the kernel 
+topology logic, this wasn't really a bug per se, but it looks like one to users.
+
+--------------------------------------------------------------------------------
+FIXES:
+
+1a. CPU: At long last, codeberg issue #307 supplied the data of a multi-die or 
+more accurately, a multi-cluster but no die_id, system. RISC-V in this case, 
+which tripped the wrong core count, and a false MT type, because it had > 1 
+cluster, but no die. Each cluster repeated the same core_ids. 
+
+That's the same issue I believe the old AMD Ryzen issue had, but I never got the 
+required data for that. Had to redo the entire /sys CPU data structure to be: 
+phyical_ids > die_ids > cluster_ids > core_ids, which was very tricky to do.
+
+1b. CPU: cpu_arch: fixed some AMD Zen generations, and process nodes, had 
+estimated from roadmap, updated with real.
+
+2a. RECOMMENDS: issue #309 notes ifconfig wasn't clearly ID'ed as:
+(deprecated, ip preferred)
+That's been corrected.
+
+2b. RECOMMENDS: issue #309 also noted that lspci is not listed, which was an
+oversight, given the equivalent BSD tools are tested for BSDs. I have no idea 
+how that was missed all these years!
+
+2c. RECOMMENDS: added ps test, found a case where that's not installed.
+
+3. GRAPHICS: GPU data: fixed an nvidia 470 Kepler2/Fermi2 ID error. And also 
+corrected the arch name, it's Kepler-2, not Kepler.
+
+--------------------------------------------------------------------------------
+ENHANCEMENTS:
+
+1. MACHINE: Added systemd.machine_id to --filter-uuid filters. Thanks Malcolm 
+from Opensuse for pointing out that one.
+
+2a. CPU: Now shows clusters, if found. Clusters seem related to L2 assignment, 
+but not always. When no die_id or cluster_id found, won't show that item.
+
+2b. CPU: New AMD, Intel CPU arch models.
+
+3a. GRAPHICS: added cosmic-comp compositor support. No current known way to get 
+version.
+
+3b. GRAPHICS: new gpu ids for intel, nvidia.
+
+4. SYSTEM: Desktop: Added basic Cosmic DE detection. Uses XDG for detection,
+and cosmic-session for possible future version.
+
+5. INFO: Power: Added system76-power power manager.
+
+6. DRIVES: a huge disk vendor update. New vendors, new matches for existing! 
+It's surreal how many SSD and USB makers there are out in the world. It's like 
+running a corner store used to be or something. There are a LOT of new matches 
+and vendors this time around!
+
+--------------------------------------------------------------------------------
+CHANGES:
+
+1. CPU: Moved the new dies:/clusters: block to right after the cpu physical 
+count, and before the core count. This makes the order of output match the order 
+of organization: physical cpus > dies > cluster > cores. Also if it got a 
+positive die_id or cluster_id, will show how many found always for -Ca. Before 
+it only showed if greater than 1 die.
+
+--------------------------------------------------------------------------------
+DOCUMENTATION:
+
+1a. MAN: Debian Lintian tests found a typo! That's fixed. I would have expected 
+many more than 1 typos, so that's good to see.
+
+1b. MAN: A missing space and extra " slightly broke the header:
+.TH INXI 1 "2024\-06-27"inxi" "inxi manual"
+should be:
+.TH INXI 1 "2024\-06-27" inxi "inxi manual"
+All that happens due to this is bottom of man output shows:
+inxi"    2024-06-18            INXI(1)
+so no big deal, but it's corrected anyway.
+
+1c. MAN: Also added inxi version to man info line, I noticed other programs do
+that. One more thing to forget during upgrades heh. Maybe I'll add that to 
+release.pl since I'm sure to forget it.
+
+1d. MAN: Updated to show clusters added, and to show cluster sample for -C a. 
+Also added more complete explanation of when dies/clusters show, and how core
+and cluster counts work.
+
+2a. OPTIONS: Fixed typo, thanks codeberg PR. Note inxi branch is not for PR, 
+only pinxi. But good to catch typos anyway.
+
+2b. OPTIONS: Added dies, clusters to -Ca, now that dies/clusters always show if 
+found, regardless of number.
+
+3a. DOCS: docs/inxi-cpu.txt: had neglected to add in the advanced sys/cpuinfo 
+pairs debugger code from the initial Slackware LQ forums threads. Added in 
+CPU DEGUGGING primary section, with that code, which is 2 1 liners, could be 1.
+
+This is same basic logic the --debug 2x debugger collects, but sometimes it's 
+easier to just get the two files we need from people.
+
+--------------------------------------------------------------------------------
+CODE:
+
+1. simplified main::filter_partitions, main::filter_pci_long. Now takes string 
+by reference, and removes all copies, so those are more efficient now. 
+
+2a. CPU: finally added in advanced topology structure phys > dies > clusters > 
+cores to the main cpu_sys data structure, I'd avoided doing that in past because 
+it's hard, and I never found a data sample that had repeated core_ids per die, 
+until this one. Note while for RISC-V, this fix applies to the entire class of 
+possible repeating core_ids per die / cluster.
+
+2b. CPU: CpuItem:: sys_data_grabber(): Cleaned up, added more globbing items for 
+/sys data. Needed to get at least cluster_id, but made globbing easier to parse 
+and read, and since the cluster stuff is useful, added more related.
+
+This fix may have unintended consequences of breaking other logic, but I believe 
+it is only within the /sys data itself.
+
+2c. CPU: refactored ordering of cpu subs, renamed many to be less ambiguous, now
+ordered more consistently. The cpu code fixes made me realize that code was not
+easy to follow in terms of where things were arranged, now it's quite 
+consistent, with clear section headers, end section etc.
+
+3. NETWORK: cleaned up code in set_ifconfig(), set_ip();
+
+4. PsData: fixed corner case where no ps is present in system. This trips some 
+undefined errors in the grabber.
+
+5. DOWNLOADER: failed to handle case of no curl OR wget, and tried running -U
+without a downloader anyway.
+
+Made the following changes:
+* Network -i: if no dig, and no downloaders, just skips grabbing IPs.
+* Updater -U: if no downloader, exit with error message.
+* Weather -w: if no downloader, will just show no connection available.
+
+6. DATA: tools/lists/gpu.nv.560.raw added. tools/gpu_raw.pl, tools/gpu_ids.pl 
+updated for nv 560 files.
+
+--------------------------------------------------------------------------------
+-- Harald Hope - Wed, 4 September 2024 10:15:08 -0700
+
+================================================================================
 Version: 3.3.35
 Patch: 00
 Date: 2024-06-18