Обсуждение: Introduce ENDLIST to terminate multiline makefile lists
Hi Hacker,
The benefits are:
This is not a new invention on my part. I’m simply proposing to bring this existing technique into the PostgreSQL source tree.
In this patch, I’ve only applied $(ENDLIST) to the Makefiles touched by [1]. If this approach is acceptable, I can follow up with a patch to apply it more broadly.
Patch [1] reminded me of a technique that was used long ago in my early C/C++ projects at Lucent Technologies (for those who still remember the company :) ). In those projects, multi-line lists in Makefiles were always terminated with $(ENDLIST), where ENDLIST is just a global dummy variable.
The benefits are:
* It clearly marks the end of a list.
* Adding or removing items does not require worrying about a trailing “\”.
* Changes at the end of a list do not introduce extra diff noise from adjusting backslashes.
In this patch, I’ve only applied $(ENDLIST) to the Makefiles touched by [1]. If this approach is acceptable, I can follow up with a patch to apply it more broadly.
Note that since ENDLIST is defined in Makefile.global.in, this change requires rerunning ./configure.
Вложения
On Mon, Dec 29, 2025 at 09:43:30AM +0800, Chao Li wrote: > Note that since ENDLIST is defined in Makefile.global.in, this change > requires rerunning ./configure. > > [1] https://postgr.es/m/DF6HDGB559U5.3MPRFCWPONEAE@jeltef.nl FWIW, I don't see an advantage in this proposal, and it has the disadvantage of more maintenance if one needs to create a new list. The current state of things is simpler. I do get the argument about C enums, but there is a difference between a grammar rule trick and something allowed in a language specification. -- Michael
Вложения
On Mon, Dec 29, 2025 at 10:08 AM Michael Paquier <michael@paquier.xyz> wrote:
On Mon, Dec 29, 2025 at 09:43:30AM +0800, Chao Li wrote:
> Note that since ENDLIST is defined in Makefile.global.in, this change
> requires rerunning ./configure.
>
> [1] https://postgr.es/m/DF6HDGB559U5.3MPRFCWPONEAE@jeltef.nl
FWIW, I don't see an advantage in this proposal, and it has the
disadvantage of more maintenance if one needs to create a new list.
The current state of things is simpler.
For example, a list is:
```
OBJS = \
a.o \
b.o
```
Now we need to add c.o, then:
```
```
OBJS = \
a.o \
b.o \ <== this line needs to add tailing \
c.o <== add this new line
```
there are two lines of changes.
With the $(ENDLIST), we just add an single line:
```
OBJS = \
a.o \
b.o \
c.o \ <== only adding this new line, a single line of diff
$(ENDLIST)
```
Could you please explain what more maintenance would be?
I do get the argument about C enums, but there is a difference between a grammar rule trick and
something allowed in a language specification.
I didn't mention C enums. How did you get that? I think I only mentioned C/C++ projects because they used Makefile for building.
Best regards,
--
Chao Li (Evan)HighGo Software Co., Ltd.
Michael Paquier <michael@paquier.xyz> writes:
> On Mon, Dec 29, 2025 at 09:43:30AM +0800, Chao Li wrote:
>> Note that since ENDLIST is defined in Makefile.global.in, this change
>> requires rerunning ./configure.
> FWIW, I don't see an advantage in this proposal, and it has the
> disadvantage of more maintenance if one needs to create a new list.
> The current state of things is simpler.
Same here. This is a new bit of notation that everyone reading our
Makefiles will need to understand, and I don't think it's worth
the extra cognitive burden.
regards, tom lane