Respect the column limit when aligning consecutive macros

[Julien-Elie]

When aligning consecutive macros, the maximum line length does not seem to be taken into account:

#define HDR_LASTCHAR_SAVE(_x)             hc[(_x)].LastChar = hc[(_x)].Value[hc[_x].Length]
#define HDR__X_ORIGINAL_NNTP_POSTING_HOST 68

which is larger than the expected ColumnLimit of 79.

Whereas the first line is properly wrapped if it is the one that causes the exceed:

#define HDR_LASTCHAR_SAVE(_xxxx) \
    hc[(_xxxx)].LastChar = hc[(_xxxx)].Value[hc[_xxxx].Length]
#define HDR__X_ORIGINAL_NNTP_POSTING_HOST 68

[Julien-Elie]

Another example:

#define MAX_MAXART_TOUT         86400 /* one day max between articles received */
#define SOFT_ARTICLE_BYTE_LIMIT (1024 * 1024 * 10) /* 10MB */

which is too long, but properly wrapped if the first line is rewritten this way:

#define MAX_MAXART_TOUT_XXXXXXX \
    86400 /* one day max between articles received */
#define SOFT_ARTICLE_BYTE_LIMIT (1024 * 1024 * 10) /* 10MB */

[mydeveloperday]

Sorry I don't understand what you are expecting? this looks correct to me.

[Julien-Elie]

As I have set ColumnLimit of 79, the following alignments are not correct. They exceed the column limit:

#define HDR_LASTCHAR_SAVE(_x)             hc[(_x)].LastChar = hc[(_x)].Value[hc[_x].Length]
#define HDR__X_ORIGINAL_NNTP_POSTING_HOST 68

#define MAX_MAXART_TOUT         86400 /* one day max between articles received */
#define SOFT_ARTICLE_BYTE_LIMIT (1024 * 1024 * 10) /* 10MB */

It looks like clang-format does not count in the resulting line length all the extra spaces it adds to smaller lines for their alignment.

[mydeveloperday]

Sorry could I ask you to be really explicit about what you see and what you are expecting?

the ] on the first line is at Column 79 so won't wrap. If I add one x to the _x it will break

and \ now on the first line is at column 79 after formatting..

Do you see something different? Did I miss understand?

[Julien-Elie]

With the following setting:

AlignConsecutiveMacros: AcrossEmptyLinesAndComments
ColumnLimit: 79

the following code remains unchanged after running clang-format on it:

#define HDR_LASTCHAR_SAVE(_x)             hc[(_x)].LastChar = hc[(_x)].Value[hc[_x].Length]
#define HDR__X_ORIGINAL_NNTP_POSTING_HOST 68

The first line ends at column 91.

I would have expected its rewrite as follows (which is not done):

#define HDR_LASTCHAR_SAVE(_x) \
    hc[(_x)].LastChar = hc[(_x)].Value[hc[_x].Length]
#define HDR__X_ORIGINAL_NNTP_POSTING_HOST 68

[mydeveloperday]

Ok thank you, now I see it... this is a tough one (not saying its not a bug)

but what should win ColumnLimit or AlignConsecutiveMacros?

[mydeveloperday]

does your opinion change if you have, i.e. should the breaking of the line break the alignment chain?

#define HDR__X_ORIGINAL_NNTP_POSTING_H  68
#define HDR__X_ORIGINAL_NNTP_POSTING_HO 68
#define HDR_LASTCHAR_SAVE(_x) \
    hc[(_x)].LastChar = hc[(_x)].Value[hc[_x].Length]
#define HDR__X_ORIGINAL_NNTP_POSTING_HOST 68

or would you expect it to be

#define HDR__X_ORIGINAL_NNTP_POSTING_H    68
#define HDR__X_ORIGINAL_NNTP_POSTING_HO   68
#define HDR_LASTCHAR_SAVE(_x) \
    hc[(_x)].LastChar = hc[(_x)].Value[hc[_x].Length]
#define HDR__X_ORIGINAL_NNTP_POSTING_HOST 68

[Julien-Elie]

Note that there is the same issue with the alignment of array of structs (#52981).

I think alignment should be done in the respect of column limit.
clang-format does the right thing if the last line is too long. It wraps it as soon as it can. For instance:

#define HDR__X_ORIGINAL_NNTP_POSTING_HOST \
    68

[Julien-Elie]

As for your example, I would expect the second one... (I see that clang-format currently does the first form but I am unsure it is what should be expected when aligning across empty lines and comments; it defeats the intent of the alignment.)

So we have 2 different issues:

• respecting column limit;
• preserving the alignment across empty lines and comments when wrapping occurred.

[Thalley]

Still an issue in version 19.1.2