Commit d36b252
committed
ci: strip stable noindex before pruning deploy noise
The prune step reverts files whose only diff versus gh-pages is build noise.
Stable pages are the canonical indexed version, so their noindex robots meta
is stripped after deploy — HEAD has no noindex, but a fresh Sphinx build does.
With the strip running after the prune, every rebuilt stable file still carried
that meta at prune time, counted as a real change, and was kept. The later
strip then removed the meta, leaving the file differing from HEAD by only the
build timestamp — exactly the noise the prune was meant to drop. Result: the
stable folder shipped hundreds of timestamp-only files (e.g. 255 in one deploy)
while 32/33/34, which are never stripped, pruned correctly.
Move the noindex strip ahead of the prune so noise-only stable pages match HEAD
and get reverted like every other version.
Assisted-by: ClaudeCode:claude-opus-4-8
Signed-off-by: skjnldsv <skjnldsv@protonmail.com>1 parent d438b25 commit d36b252
1 file changed
Lines changed: 12 additions & 9 deletions
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
178 | 178 | | |
179 | 179 | | |
180 | 180 | | |
| 181 | + | |
| 182 | + | |
| 183 | + | |
| 184 | + | |
| 185 | + | |
| 186 | + | |
| 187 | + | |
| 188 | + | |
| 189 | + | |
| 190 | + | |
| 191 | + | |
| 192 | + | |
181 | 193 | | |
182 | 194 | | |
183 | 195 | | |
| |||
237 | 249 | | |
238 | 250 | | |
239 | 251 | | |
240 | | - | |
241 | | - | |
242 | | - | |
243 | | - | |
244 | | - | |
245 | | - | |
246 | | - | |
247 | | - | |
248 | | - | |
249 | 252 | | |
250 | 253 | | |
251 | 254 | | |
| |||
0 commit comments