Why .SECONDARY does not work with patterns (%) while .PRECIOUS does?

Question

My question is to understand better what i missed in make process and .SECONDARY purpose vs .PRECIOUS, not to get my script working, since it does work already.

I am using make to either open a emacs editor on a file ( java but irrelevant for purpose of this question ) or to create it with a template if not existing.

If it works well with existing files, when using generated file it is removed at the end.

I added prerequisite in .SECONDARY but didn't help, i had to add it in .PRECIOUS.

This is question why wasn't it working in .SECONDARY ? .

From what i found on SO .SECONDARY does not work with patterns ( % ), but even with knowing that i wonder if it is by design or if it is a bug in make. ( .SECONDARY for a pattern rule with GNU Make and Makefile pattern rule either ignores phony rule or spontaneously deletes output file )

Here a stripped down content of my Makefile to reproduce my problem ( please create a com/stackoverflow/question directory to test it ).

PACKAGE=com.stackoverflow.question
PACKAGE_DIR=$(subst .,/,$(PACKAGE))
OUT=out

clean:
    find $(OUT) -name "*.class" -type f -print0|xargs -0 rm

# does not work : deleted at end due to intermediate file removal.
$(PACKAGE_DIR)/%.java:
    @echo "package com.stackoverflow.question;\npublic class $(subst .java,,$(subst $(PACKAGE_DIR)/,,$@))\n{\n /** TODO */ \n}" >$@ 

work/%: $(PACKAGE_DIR)/$(subst work/,,%).java
    emacs $<

.PHONY: clean work/%

# tried to avoid intermediate file removal : does not work
.SECONDARY: $(PACKAGE_DIR)/%.java 

# if not commented this does work : once precious intermediate file is not removed.
#.PRECIOUS: $(PACKAGE_DIR)/%.java 

try

make work/SoTest

I understand this is flagged intermediate.

then looking in SO i tried to set it in .SECONDARY: target list : does not work either.

looking at make source code i spotted that make intermediate files removal is done within this context :

if (f->intermediate && (f->dontcare || !f->precious)
    && !f->secondary && !f->cmd_target)

so i set my file in .PRECIOUS: and now it works.

it displays to console :

com/stackoverflow/question/SoTest.java

it run emacs with right template in it so creation is OK here i exit emacs

and it removes the file at the end

rm com/stackoverflow/question/SoTest.java

Removal at end is due to intermediate file, this can be seen with -d option on make

LANG=C make -d work/SoTest

...
Must remake target 'work/SoTest'.
emacs com/stackoverflow/question/SoTest.java
Putting child 0xc3b580 (work/SoTest) PID 20681 on the chain.
Live child 0xc3b580 (work/SoTest) PID 20681 
Reaping winning child 0xc3b580 PID 20681 
Removing child 0xc3b580 PID 20681 from chain.
Successfully remade target file 'work/SoTest'.
Removing intermediate files...
rm com/stackoverflow/question/SoTest.java

To have it working i need to uncomment the .PRECIOUS paragraph.

make --version

GNU Make 4.0
Construit pour x86_64-pc-linux-gnu
Copyright (C) 1988-2013 Free Software Foundation, Inc.
Licence GPLv3+ : GNU GPL version 3 ou ultérieure <http://gnu.org/licenses/gpl.html>
Ceci est un logiciel libre : vous êtes autorisé à le modifier et à la redistribuer.
Il ne comporte AUCUNE GARANTIE, dans la mesure de ce que permet la loi.

Answer 1

Thanks to Alex (see answer) i went further in my search.

What i found is that it is recorded in TODO.private of make project for 15 years ....

Using git://git.savannah.gnu.org/make.git you can see history of TODO.private content :

 6) Right now the .PRECIOUS, .INTERMEDIATE, and .SECONDARY
    pseudo-targets have different capabilities.  For example, .PRECIOUS
    can take a "%", the others can't.  Etc.  These should all work the
    same, insofar as that makes sense.

These should all work the same, insofar as that makes sense. but was not coded.

Answer 2

The answer to "Why .SECONDARY does not work with patterns (%) while .PRECIOUS does?" is here: the document says

You can also list the target pattern of an implicit rule (such as ‘%.o’) as a prerequisite file of the special target .PRECIOUS

but does not say this about .SECONDARY. But for the few explicit exceptions, none of the special targets accept patterns.