Commit bce5ff91 authored by Glenn Morris's avatar Glenn Morris
Browse files

; * lisp/Makefile.in: Comments about loaddefs.

parent 8f2062d9
......@@ -173,31 +173,46 @@ $(lisp)/finder-inf.el:
--eval '(setq generated-finder-keywords-file (unmsys--file-name "$(srcdir)/finder-inf.el"))' \
-f finder-compile-keywords-make-dist ${SUBDIRS_FINDER}
# Use expand-file-name rather than $abs_scrdir so that Emacs does not
# get confused when it compares file-names for equality.
#
# Note that we set no-update-autoloads in _generated_ leim files.
# If you want to allow autoloads in such files, remove that,
# and make this depend on leim.
# Actually this should depend on leim anyway, since no-update-autoloads
# files are still recorded in loaddefs. So we should ensure
# that all input files are generated before we create loaddefs.
# Otherwise making loaddefs again will change the output.
#
# In fact, now we rely on this target to create leim-list.
# In src, emacs depends directly on loaddefs.el (not leim-list).
## Comments on loaddefs generation:
# loaddefs depends on gen-lisp for two reasons:
# 1) In ../src, the emacs target depends on loaddefs but not on eg leim-list.
# So having leim as a dependency of loaddefs (via gen-lisp) ensures leim-list
# gets created before the final emacs is dumped. Having leim
# dependencies in ../src as well would create a parallel race condition.
#
# Write to a temporary file in case we're doing a parallel build and a
# CANNOT_DUMP-mode Emacs needs to read loaddefs at startup.
# 2) Files that are marked no-update-autoloads still get recorded in loaddefs.
# So those files should be generated before we make autoloads, if we
# don't want a successive make autoloads to change the output file.
# Said changes are trivial (only comments in the "files without autoloads"
# section), but still can be annoying. Of course, if generated lisp files
# do contain autoloads, it is essential they be built before make autoloads.
# (Also, if a generated file is not written atomically, it is possible that
# in a parallel build, make autoloads could read a partial version of it.)
#
# We'd really like to add "make -C ../admin/unidata all" to gen-lisp
# because of 2) above, but it causes a race condition in parallel
# builds because ../src also runs that rule. Given the limitations of
# recursive make, the only way to fix that would be to remove unidata
# from ../src rules, but that doesn't seem possible due to the various
# non-trivial dependencies.
# We make $(lisp)/loaddefs.el a dependency of .PHONY to cause Make to
# ignore its time stamp. That's because the real dependencies of
# loaddefs.el aren't known to Make, they are implemented in
# batch-update-autoloads, which only updates the autoloads whose
# sources have changed. We start by copying an existing loaddefs.el
# to loaddefs.tmp to avoid regenerating the entire file anew, which is
# slow; starting from an almost-correct content will enable the "only
# update where necessary" feature of batch-update-autoloads.
# sources have changed.
# Write to a temporary file in case we're doing a parallel build and a
# CANNOT_DUMP-mode Emacs needs to read loaddefs at startup. We start
# by copying an existing loaddefs.el to loaddefs.tmp to avoid regenerating
# the entire file anew, which is slow; starting from an almost-correct
# content will enable the "only update where necessary" feature of
# batch-update-autoloads.
# Use expand-file-name rather than $abs_scrdir so that Emacs does not
# get confused when it compares file-names for equality.
autoloads .PHONY: $(lisp)/loaddefs.el
$(lisp)/loaddefs.el: gen-lisp $(LOADDEFS)
@echo Directories for loaddefs: ${SUBDIRS_ALMOST}
......@@ -352,6 +367,8 @@ compile-clean:
.PHONY: gen-lisp leim semantic
## make -C ../admin/unidata all should be here, but that would race
## with ../src. See comments above for loaddefs.
gen-lisp: leim semantic
leim:
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment