Message ID | cover-v2-0.4-00000000000-20210824T091204Z-avarab@gmail.com (mailing list archive) |
---|---|
Headers | show |
Series | test-tool: split up "read-cache" tool | expand |
On Tue, Aug 24, 2021 at 11:15:21AM +0200, Ævar Arnfjörð Bjarmason wrote: > A re-roll addressing the feedback v1 got, see > https://lore.kernel.org/git/cover-0.4-0000000000-20210607T115454Z-avarab@gmail.com > > I think the gist of that feedback from Emily was that v1 could be > understood as aiming to optimize the perf test, but the main reason > I'm doing this split up is to make the code easier to read. The > changed commit message summarizes both goals. > > I also changed some nits in the code I spotted myself, e.g. "argc > 0" > checks to just "argc", and a simpler way of providing the "cnt" > default. Playing devil's advocate for a moment: I think that the current implementation is somewhat easier to read as it lives in a single file, and makes clear that each option implies different behavior through the body of the loop. If we're looking for things to clean up, I do like the conversion to the parse-options API instead of reading argv ourselves, but probably otherwise prefer the code as-is instead of split across many files. But I may be in the minority, and there may be others who do find the split-up version easier to grok. Thanks, Taylor
Taylor Blau <me@ttaylorr.com> writes: > If we're looking for things to clean up, I do like the conversion to the > parse-options API instead of reading argv ourselves, but probably > otherwise prefer the code as-is instead of split across many files. > > But I may be in the minority, and there may be others who do find the > split-up version easier to grok. FWIW, you're not alone. I too like the use of parse_options, but I fail to be enthused by changes to churn the test helper binary.
On Tue, Aug 24 2021, Junio C Hamano wrote: > Taylor Blau <me@ttaylorr.com> writes: > >> If we're looking for things to clean up, I do like the conversion to the >> parse-options API instead of reading argv ourselves, but probably >> otherwise prefer the code as-is instead of split across many files. >> >> But I may be in the minority, and there may be others who do find the >> split-up version easier to grok. > > FWIW, you're not alone. I too like the use of parse_options, but I > fail to be enthused by changes to churn the test helper binary. I think it's an improvement as-is & would like you to pick it up, but I understand if you disagree. This series converts the code to parse_options(), but also changes the perf test to not load config, and adds a --refresh option for a new perf test. I tried to make that readable all in one function, but I think it just ends up being more of a confusing if/else jungle than the current state. I don't think I'm the right person to re-roll this in that way & argue for it being an improvement when I don't think it would be.
A re-roll addressing the feedback v1 got, see https://lore.kernel.org/git/cover-0.4-0000000000-20210607T115454Z-avarab@gmail.com I think the gist of that feedback from Emily was that v1 could be understood as aiming to optimize the perf test, but the main reason I'm doing this split up is to make the code easier to read. The changed commit message summarizes both goals. I also changed some nits in the code I spotted myself, e.g. "argc > 0" checks to just "argc", and a simpler way of providing the "cnt" default. Ævar Arnfjörð Bjarmason (4): test-tool: split up test-tool read-cache test-tool: migrate read-cache-perf to parse_options() test-tool: migrate read-cache-again to parse_options() read-cache perf: add a perf test for refresh_index() Makefile | 2 ++ t/helper/test-read-cache-again.c | 45 +++++++++++++++++++++++++ t/helper/test-read-cache-perf.c | 47 +++++++++++++++++++++++++++ t/helper/test-read-cache.c | 56 +++++++++++++------------------- t/helper/test-tool.c | 2 ++ t/helper/test-tool.h | 2 ++ t/perf/p0002-read-cache.sh | 7 +++- t/t7519-status-fsmonitor.sh | 2 +- 8 files changed, 128 insertions(+), 35 deletions(-) create mode 100644 t/helper/test-read-cache-again.c create mode 100644 t/helper/test-read-cache-perf.c Range-diff against v1: 1: 6e7fcd46934 ! 1: adb3f989a29 test-tool: split up test-tool read-cache @@ Commit message test-tool: split up test-tool read-cache Since the "test-tool read-cache" was originally added back in - 1ecb5ff141 (read-cache: add simple performance test, 2013-06-09) it's - been growing all sorts of bells and whistles that aren't very - conducive to performance testing the index, e.g. it learned how to - read config. + 1ecb5ff141 (read-cache: add simple performance test, 2013-06-09) the + test-read-cache.c tool has been growing various features that make the + code harder to read. I.e. sometimes running as a one-off, sometimes looping. + + It's also been unconditionally reading config since + dc76852df2f (fsmonitor: demonstrate that it is not refreshed after + discard_index(), 2019-05-07), which introduces unnecessary noise into + the performance test. Then in recent changes in e2df6c3972 (test-read-cache: print cache entries with --table, 2021-03-30) and 2782db3eed (test-tool: don't @@ Commit message I think that having one test tool do so many different things makes it harder to read its code. Let's instead split up the "again" and "perf" - uses for it into their own tools. + uses for it into their own small tools, this makes the main + "test-read-cache.c" a simpler. Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> @@ t/helper/test-read-cache.c: static void print_cache(struct index_state *istate) + }; + + argc = parse_options(argc, argv, "test-tools", options, read_cache_usage, 0); -+ if (argc > 0) ++ if (argc) + usage_msg_opt("Too many arguments.", read_cache_usage, options); initialize_the_repository(); 2: 07f392e0878 ! 2: a68fa4a6355 test-tools: migrate read-cache-perf to parse_options() @@ Metadata Author: Ævar Arnfjörð Bjarmason <avarab@gmail.com> ## Commit message ## - test-tools: migrate read-cache-perf to parse_options() + test-tool: migrate read-cache-perf to parse_options() Change the newly added (but then mostly copy/pasted) read-cache-perf to use the parse_options() API. This will make things easier as we add @@ t/helper/test-read-cache-perf.c - die("usage: test-tool read-cache-perf [<count>]"); + argc = parse_options(argc, argv, "test-tools", options, + read_cache_perf_usage, 0); -+ if (argc > 0) ++ if (argc) + usage_msg_opt("Too many arguments.", read_cache_perf_usage, + options); + if (cnt < 1) 3: 36f4072b131 ! 3: a34e69eaa48 test-tools: migrate read-cache-again to parse_options() @@ Metadata Author: Ævar Arnfjörð Bjarmason <avarab@gmail.com> ## Commit message ## - test-tools: migrate read-cache-again to parse_options() + test-tool: migrate read-cache-again to parse_options() Change the newly added (but then mostly copy/pasted) read-cache-perf to use the parse_options() API. I have no plans to further modify @@ t/helper/test-read-cache-again.c { struct repository *r = the_repository; - int i, cnt; -+ int cnt = -1; ++ int cnt = 2; const char *name; + struct option options[] = { + OPT_INTEGER(0, "count", &cnt, "number of passes"), @@ t/helper/test-read-cache-again.c + if (argc != 1) + usage_msg_opt("Too many arguments.", read_cache_again_usage, + options); -+ if (cnt == -1) -+ cnt = 2; -+ else if (cnt < 1) ++ if (cnt < 1) + usage_msg_opt("Need at least one pass.", read_cache_again_usage, + options); name = argv[2]; 4: 120a37acaef = 4: e3648bf78c7 read-cache perf: add a perf test for refresh_index()