From: Max Kirillov <max@xxxxxxxxxx> Subject: http-backend test: make empty CONTENT_LENGTH test more realistic This is a test of smart HTTP, so it should use the smart HTTP endpoints (e.g. /info/refs?service=git-receive-pack), not dumb HTTP (HEAD). Signed-off-by: Max Kirillov <max@xxxxxxxxxx> Signed-off-by: Jonathan Nieder <jrnieder@xxxxxxxxx> --- Max Kirillov wrote: > Provided more thorough message, also fix test (it did not test > actually the error before) > > There will be more versions later, at least the one which suggested > by Jeff v2 is in "next", and I believe that version should already be sufficient for Git 2.19. Please correct me if I'm wrong. Since v2 is in "next", I think any further refinements are supposed to be incremental patches on top. Here's an example (representing the v2->v3 diff). It's more of an RFC than a serious patch, because: This version of the test doesn't seem to reproduce the bug. When I run the test against the unfixed version of http-backend, it passes. Ideas? Not about this patch: could this test share some infrustructure with t5560-http-backend-noserver.sh? If there were some common shell library that they shared, the tests might be easier to read and write. Thanks, Jonathan t/t5562-http-backend-content-length.sh | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/t/t5562-http-backend-content-length.sh b/t/t5562-http-backend-content-length.sh index f94d01f69e..fceb3d39c1 100755 --- a/t/t5562-http-backend-content-length.sh +++ b/t/t5562-http-backend-content-length.sh @@ -155,8 +155,8 @@ test_expect_success 'CONTENT_LENGTH overflow ssite_t' ' test_expect_success 'empty CONTENT_LENGTH' ' env \ - QUERY_STRING=/repo.git/HEAD \ - PATH_TRANSLATED="$PWD"/.git/HEAD \ + QUERY_STRING="/repo.git/info/refs?service=git-receive-pack" \ + PATH_TRANSLATED="$PWD"/.git/info/refs \ GIT_HTTP_EXPORT_ALL=TRUE \ REQUEST_METHOD=GET \ CONTENT_LENGTH="" \ -- 2.19.0.rc2.392.g5ba43deb5a
