[FlareCoding]

Summary

The display engine's set-up printed a count of 1 beside a plural word, as in "1 timings", "EDID of 1 bytes", "up to 1 lanes" and "1 symbols of horizontal blanking", in lines counting the DisplayPort outputs, a link's lanes, the rows of the base buffer (the video memory the display engine's window is to scan out), the symbols (each a byte of pixels, other data or framing) that horizontal blanking leaves free in a link's stream, or a monitor's EDID (the monitor's description of itself) by its timings, bytes or skipped extension blocks (its 128-byte blocks after the first).

Changes

• Each counted word follows its count: log_monitor (display_engine.cpp) chooses "timing" or "timings" and "block" or "blocks", log_connector_link_limits and log_monitor_link_limits "lane" or "lanes", and the mode line in keep_output_modes "timing" or "timings", each by the count printed beside it. configure_stream (display_link.cpp) chooses "symbol" or "symbols" of horizontal blanking.
• Lines reworded to read right at any count: read_display_facts logs "the static digital display ids give 3 DisplayPort outputs" and set_up_base_buffer "its 1440 rows of 10240 bytes", each word chosen as above. The lines in keep_output_modes giving an output no mode read "output 0 has no mode, as its EDID gives 2 timings and its link carries none", with " below 60.600 Hz" ending the one for an output limited to 60 Hz.
• A refused EDID is named by its size: read_once (display_edid.cpp) logs an EDID edid::parse refuses as in "display id 0x100's 127-byte EDID is shorter than one block", the form display_dpcd.cpp's lines use for the size of a read from the monitor.
• Card: on the RTX 3080 driving an IOC 25G2S monitor over DisplayPort, run land-display-singular-s4-1, a boot of 592cb760, an earlier version whose kernel, stripped of debug information, is byte-identical to this commit's, logs "the static digital display ids give 3 DisplayPort outputs", the three connectors' and the monitor's limits of 4 lanes, the monitor's 6 timings with 0 extension blocks skipped, the mode line's 6 timings, the base buffer's 1440 rows of 10240 bytes and the stream's 499 symbols of horizontal blanking, each in the plural, with the mode line's "over 1 lane" in the singular. Bring-up ends with the driver attached, and the host reports 0 IOMMU or PCIe faults.
• Tests of the singular: nvidia_display_engine checks the outputs line, both link-limit lines, the monitor's line and the mode line on one output with a link of one lane at each end and an EDID giving one timing and counting one extension block it does not hold, and nvidia_display_link checks the stream line at 1 symbol of horizontal blanking and at 0. The engine suite also checks the mode line of an output limited to 60 Hz keeping one of two timings, the lines giving no mode when one lane at RBR (DisplayPort's lowest link rate) cannot carry the one timing an EDID gives, with and without that limit, and the base buffer's line for a mode one row high.
• Tests of the plural and of each word's own count: on two outputs with a link of four lanes at each end and an EDID giving two timings and counting two extension blocks, nvidia_display_engine checks the plurals of the outputs line, the monitor's line and the link-limit lines, and on an EDID giving one timing and counting two extension blocks it checks that each counted word of the monitor line follows its own count.
• Tests of the reworded lines: nvidia_display_engine checks the lines giving no mode at 2 timings and the base buffer's line at 768 rows as reworded, and nvidia_display_edid checks the refused EDID's line at sizes of 0, 127 and 128 bytes.

[cursor]

Current version of PR was reviewed by /review-bugbot with additional instructions on Oct 29, 19:57. It flagged 0 findings.

^{Bugbot on commit dd2dabe is skipped.}