mirror of
git://git.sv.gnu.org/autoconf
synced 2025-01-06 10:25:53 +08:00
5ccf45a370
The general idea is this: all write failures from `configure' writing `config.status' are indicated by $ac_write_error, which is only checked at the end. This is safe because config.status code is not executed before the file is complete. Other write failures, be they inside config.status, or in sub shell/awk scripts spawned from configure or config.status, typically need earlier checking, as their results are used right afterwards. * lib/autoconf/status.m4 (AC_OUTPUT): Initialize `ac_write_fail' before writing config.status, check afterwards. (_AC_OUTPUT_FILES_PREPARE, _AC_OUTPUT_FILE) (_AC_OUTPUT_HEADERS_PREPARE,_AC_OUTPUT_CONFIG_STATUS): Set `ac_write_error' for write failures to config.status. Barf upon write failures to temporary files. Adjust note about closing and reopening the here-document. (_AC_OUTPUT_HEADER, _AC_OUTPUT_LINK, _AC_OUTPUT_COMMAND) (_AC_OUTPUT_MAIN_LOOP): Likewise, adjust note about closing and reopening the here-document. * tests/torture.at (AC_CONFIG_FILES, HEADERS, LINKS and COMMANDS): Ensure `ac_write_error' does not escape into config.status. Also, add a couple of code paths not yet exercised in the test suite: a config file with input from stdin, and a config header output to stdout. Suggestion for catching write errors by Bruno Haible. |
||
|---|---|---|
| .. | ||
| autoconf | ||
| Autom4te | ||
| autoscan | ||
| autotest | ||
| emacs | ||
| m4sugar | ||
| .cvsignore | ||
| .gitignore | ||
| autom4te.in | ||
| freeze.mk | ||
| Makefile.am | ||