Hi, I'm not sure if this is the right list, if not please redirect me. So, I tried compiling luametatex using g++, and noticed some things that prevent it from also being considered valid C++ code. Two general problems are: 1) Initialization order. C++ wants initializers to be specified in the order in which they will actually initialize the object, i.e. in the order of the members. 2) Reserved names. Some variables are named `class` or `template`, which of course, breaks C++. But more interesting are the specific pieces of code that get rejected, because C++ is a bit more strict than C in some cases. 3) We get one complaint about comparing a pointer to an integer using <, which I believe is an actual bug in the code. 4) A second complaint is about dropping const in `tex_to_cstring`. I don't think this causes an actual error in the program, though. 5) Some of the `goto`s cause problems, because the jump skips over the initialization of variables in the target scope. Again, these variables are not used, so I don't think there is an actual error. The easy fix here is to just move the code from the `goto` target into its own function, which is probably better for readability anyway. 6) The last problem is that things break with g++ if the marco `infinity` is defined. I've attached patches for 3), 4), and 5), as I think that these benefit code quality in general. I can also send ones for 1) and 2), if there is interest. 6) seems to be a more general incompatibility, I'm not sure if there is a "fix" that doesn't require #if s. Kind regards, Erik
On 12/12/2022 2:11 PM, Schultheis Erik wrote:
I'm not sure if this is the right list, if not please redirect me.
this list or the general context list
So, I tried compiling luametatex using g++, and noticed some things that prevent it from also being considered valid C++ code.
ok, but keep in mind that we try to avoid c++ specifics
Two general problems are:
1) Initialization order. C++ wants initializers to be specified in the order in which they will actually initialize the object, i.e. in the order of the members.
Do you have a list of locations?
2) Reserved names. Some variables are named `class` or `template`, which of course, breaks C++.
i'll replace these (but you need to check it as i don't compile with g++
But more interesting are the specific pieces of code that get rejected, because C++ is a bit more strict than C in some cases.
interesting; we compile with clang, gcc and msvc and each has its own (sometimes conflicting) messages
3) We get one complaint about comparing a pointer to an integer using <, which I believe is an actual bug in the code.
ok, fixed
4) A second complaint is about dropping const in `tex_to_cstring`. I don't think this causes an actual error in the program, though.
hm, changes but let's see what othjer compilers say
5) Some of the `goto`s cause problems, because the jump skips over the initialization of variables in the target scope. Again, these variables are not used, so I don't think there is an actual error. The easy fix here is to just move the code from the `goto` target into its own function, which is probably better for readability anyway.
adapted (slightly different)
6) The last problem is that things break with g++ if the marco `infinity` is defined.
adapted but best check if i didn't make mistakes there
I've attached patches for 3), 4), and 5), as I think that these benefit code quality in general. I can also send ones for 1) and 2), if there is interest. 6) seems to be a more general incompatibility, I'm not sure if there is a "fix" that doesn't require #if s.
you can check (1) and (2) next upload; we really try to avoid compiler specific #if's Hans ----------------------------------------------------------------- Hans Hagen | PRAGMA ADE Ridderstraat 27 | 8061 GH Hasselt | The Netherlands tel: 038 477 53 69 | www.pragma-ade.nl | www.pragma-pod.nl -----------------------------------------------------------------
participants (2)
-
Hans Hagen
-
Schultheis Erik