make2graph output discrepancy if file already exists

[romangrothausmann]

Considering this simple Makefile:

%.a:
echo "$@" > $@

%.b : %.a
echo "$@" > $@

%.z : %.b
echo "$@" > $@

If 1.a does not exist make2graph yields:
rm 1.a
/opt/make-4.1/bin/make -Bnd 1.z | make2graph | dot -Tpng -o Makefile_1.png

However, if 1.a already exists an additional (false?) dependency is drawn:
touch 1.a
/opt/make-4.1/bin/make -Bnd 1.z | make2graph | dot -Tpng -o Makefile_2.png

Same problem with make-3.81 or j4make

[lindenb]

I quickly looked at the following Makefile

%.b: %.a
    touch $@

%.z:%.b
    touch $@

%.a:
    touch $@

all: x.z

I confirme I get the same output as you, but the output of Make in debug mode shows me a depency. Pruning filex.a'.`
I think it comes from the fact that you're using pattern rules.

 make -nrd | grep -vE '(Considering |Found|Trying pattern|Trying impli|Looking for|Avoiding im|Finished pre)'

output:

Updating goal targets....
 File `all' does not exist.
   File `x.z' does not exist.
     No need to remake target `x.a'.
    File `x.b' does not exist.
     Pruning file `x.a'.
   Must remake target `x.b'.
touch x.b
   Successfully remade target file `x.b'.
  Must remake target `x.z'.
touch x.z
  Successfully remade target file `x.z'.
Must remake target `all'.
Successfully remade target file `all'.
Removing intermediate files...
rm x.b

sorry, there is not much I can do.

P.

[romangrothausmann]

Hm, would this then be a bug in the debug output of make?
Would it work to ignore any "Pruning file XYZ" if a "No need to remake target XYZ" appears before on the same dependency level?

[lindenb]

Hm, would this then be a bug in the debug output of make?

yes and no, make is not supposed to know I want to use the information :-)

Would it work to ignore any "Pruning file XYZ" if a "No need to remake target XYZ" appears before on the same dependency level?

no that would break everything.

[romangrothausmann]

I just noticed that if the files make would create actually exist, the visualization/"make -nd output" is correct. E.g. add to the Makefile a line with:
.SECONDARY

run make without -n to create all files
make 1.z
and then run again:
make -Bnd 1.z | make2graph | dot -Tpng -o Makefile_3.png

[romangrothausmann]

Using make -Bt first lets make2graph work correctly even without .SECONDARY in the Makefile, as make does not remove intermediate files that were crated by -t.

[deryni]

Having looked at this for only a few minutes (and not having looked at the actual code for make2graph yet) it looks like (from the make -rRd output of this makefile in the two states that it should be possible to match up the prerequisite with the correct target.

The output indented under the processing of x.b stops at Trying implicit prerequisite 'x.a' in the partial case (where it continues in the clean case).

In the partial case the Considering target file 'x.a' output then appears under the Found an implicit rule for 'x.z' line but at the indentation level of the x.b lines from earlier (and not at the indentation level of the Considering target file 'x.b' lines that follow it shortly.

Similarly the No need to remake target 'x.a'. line in the output above is indented at the x.b prerequisite level and not at the x.z prerequisite level.

I don't know if that's enough to be canonical or reliable but it looks (from this limited subset of outputs) that it might be.

[romangrothausmann]

Thanks Etan for looking into this and Your suggestion to regard the indentation level.
As far as I understand the code of make2graph.c, it does not take into account the indentation level but seems to strip space (https://github.com/lindenb/makefile2graph/blob/master/make2graph.c#L191).
Pierre, how difficult would You think it will be to adjust the code to take indentation levels into account?

[lindenb]

Pierre, how difficult would You think it will be to adjust the code to take indentation levels into account?

I don't know, it shouldn't be difficult, but honestly I don't currently have the time for this.

[lindenb]

Looks nice now