)]}'
{"/PATCHSET_LEVEL":[{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"ee3ba603c306c8378c41ca09e544043186609a94","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":1,"id":"51096fed_9ccc4132","updated":"2026-09-01 14:01:53.000000000","message":"It\u0027s a hack and not the solution. I would rather go for implementing a proper auth plugin for ec2creds. This would allow having proper flexibility with disabling the whole ec2creds world with just one config change, do not require a hack which claims not to cause collision, but that is only partially true and it depends on the int size of the python what may or may not change depending on python version, architecture, etc.\nI acknowledge some clouds may went with this change during the bug disclosure, but we should stop accepting workarounds into Keystone and rather focus on proper solutions. With that the proposed change will help in the meanwhile for those who adopted it, but it would be (hopefully) overwritten with a proper solution.","commit_id":"dbb5c174705153f23f93d143c951a2008c9912b6"},{"author":{"_account_id":34637,"name":"Milana Levy","email":"millevy@redhat.com","username":"millevy"},"change_message_id":"00ce3339b0a50425ede62d22a3f776a5e9c1fecd","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":1,"id":"1c762c01_a0564be6","updated":"2026-11-09 12:08:45.000000000","message":"the idea is right, but the change is made in the wrong function, and on current master it breaks ec2credential.\n\n1. It adds names that are not in the config.\nconstruct_method_map_from_config() should return only the methods from CONF.auth.methods. This patch always adds ec2credential and oauth2_credential, even when they are not configured.\n\nWhy this is a problem:\n\nThe function no longer does what its name says.\nThe tests had to be changed to copy the private _PSEUDO_METHOD_BITS dict, so they no longer check the real result.\nThe config is no longer the only source of the bitmask.\n2. On master, ec2credential now has two bits.\nec2credential is already a registered method (0e21667), so it already has bit 64. The patch adds a second bit (1\u003c\u003c32) for the same name. I ran the code on master:\n\n\nconvert_method_list_to_integer([\u0027ec2credential\u0027]) \u003d 0x100000040   (both bits)\nconvert_integer_to_method_list(0x100000040)       \u003d [\u0027ec2credential\u0027, \u0027ec2credential\u0027]\nThe name comes back twice. Existing tests fail: test_default_method_map_preserves_legacy_bits, test_ec2_token_method_survives_fernet_round_trip, test_s3_token_method_survives_fernet_round_trip.\n\n3. The reserved bits collide with normal methods.\nNormal methods get bits 1, 2, 4, 8, ... The 33rd method gets 1\u003c\u003c32 — the same bit reserved for ec2credential. method_map.update() overwrites it, so the 33rd method disappears. I tested this:\n\n\n33 methods, last one is \u0027custom_33rd\u0027\nconvert_method_list_to_integer([\u0027custom_33rd\u0027]) \u003d 0\nconvert_integer_to_method_list(0)               \u003d []\nThere is no limit on [auth] methods today.\n\nWhat I suggest:\n\nmaster: do not merge this. Make the two names real auth methods instead, like ec2credential in 0e21667 and the proposal for oauth2_credential in 1005070. Then the map only contains config methods, and there are no reserved bits.\nstable backports (1002325, 1002326): the workaround is acceptable, because a new auth method cannot be backported. But keep construct_method_map_from_config() unchanged, add the two bits in the encode/decode functions, check for conflicts, and add tests for 33 methods and for the default config.","commit_id":"dbb5c174705153f23f93d143c951a2008c9912b6"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"45d25409db2599af583b8b2643f6bc8dab56cfbd","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":1,"id":"cab837e0_be5e6b7a","in_reply_to":"45d2ecee_50a29b94","updated":"2026-09-02 11:24:36.000000000","message":"neah, I propose not to merge it at all, at least not in master","commit_id":"dbb5c174705153f23f93d143c951a2008c9912b6"},{"author":{"_account_id":14250,"name":"Grzegorz Grasza","email":"xek@redhat.com","username":"xek"},"change_message_id":"51e7d460cbfa589a68a67e62ec4f9775d40dfa6e","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":1,"id":"45d2ecee_50a29b94","in_reply_to":"51096fed_9ccc4132","updated":"2026-09-02 11:22:47.000000000","message":"So you propose merging it now, but work on making both \"native\" auth methods?","commit_id":"dbb5c174705153f23f93d143c951a2008c9912b6"}],"keystone/auth/plugins/core.py":[{"author":{"_account_id":13252,"name":"Dr. Jens Harbott","display_name":"Jens Harbott (frickler)","email":"frickler@offenerstapel.de","username":"jrosenboom"},"change_message_id":"8dc6ab71ecc2497122ea2f411b3c040943ab0267","unresolved":true,"context_lines":[{"line_number":50,"context_line":"    method_index \u003d 1"},{"line_number":51,"context_line":"    for method in CONF.auth.methods:"},{"line_number":52,"context_line":"        method_map[method_index] \u003d method"},{"line_number":53,"context_line":"        method_index \u003d method_index * 2"},{"line_number":54,"context_line":""},{"line_number":55,"context_line":"    method_map.update(_PSEUDO_METHOD_BITS)"},{"line_number":56,"context_line":""}],"source_content_type":"text/x-python","patch_set":1,"id":"808ad0ff_4a2d1aa3","line":53,"updated":"2026-08-27 10:29:11.000000000","message":"wouldn\u0027t it be better to make sure there are at most 32 methods configured?","commit_id":"dbb5c174705153f23f93d143c951a2008c9912b6"}]}
