argp: argp.doc prints incorrectly when it starts with "\v" [BZ #19038]

Message ID CALkY8p960_gyY_Q1C+FPKuRYYyBCyp7V=PZpu8XaNW5Mxkxdgg@mail.gmail.com
State New
Headers show
Series
  • argp: argp.doc prints incorrectly when it starts with "\v" [BZ #19038]
Related show

Commit Message

Girish Joshi May 6, 2019, 8:55 a.m.
In argp-help.c `char *vt` is being initialized only once. It needs to
be initialized for every child in the doc and it needs to be printed
only in two cases.
1. There are no children and the doc does not starts with '\v'.
2. Argument `first_only` to the function `argp_doc` is false and the
complete doc needs to be printed.

     }
@@ -1498,8 +1499,11 @@ argp_doc (const struct argp *argp, const struct
argp_state *state,

       if (text == inp_text && inp_text_limit)
        __argp_fmtstream_write (stream, inp_text, inp_text_limit);
-      else
-       __argp_fmtstream_puts (stream, text);
+      else{
+        if((!vt && !child) || (text == inp_text && !first_only)){
+          __argp_fmtstream_puts (stream, text);
+       }
+      }

       if (__argp_fmtstream_point (stream) > __argp_fmtstream_lmargin (stream))
        __argp_fmtstream_putc (stream, '\n');

Comments

Florian Weimer May 6, 2019, 12:37 p.m. | #1
* Girish Joshi:

> In argp-help.c `char *vt` is being initialized only once. It needs to

> be initialized for every child in the doc and it needs to be printed

> only in two cases.


> 1. There are no children and the doc does not starts with '\v'.

> 2. Argument `first_only` to the function `argp_doc` is false and the

> complete doc needs to be printed.


I assume this bug user-visible, so it should have a bug in Bugzilla
<https://sourceware.org/bugzilla/>.  Would you please file one?

Could you write a test case for this?  Or at least post how to reproduce
the issue in a minimal program?

Thanks,
Florian
Girish Joshi May 6, 2019, 1:38 p.m. | #2
Hi Florian,

> I assume this bug user-visible, so it should have a bug in Bugzilla

> <https://sourceware.org/bugzilla/>.  Would you please file one?

>

Here[1] is a Bugzilla entry for this.

> Could you write a test case for this?  Or at least post how to reproduce

> the issue in a minimal program?

>

It can be reproduced with following code.

#include<stdlib.h>
#include<argp.h>

static char doc[] = "\vthis is post_doc";
static struct argp argp = {NULL, NULL, NULL, doc};

int main(int argc, char *args[]){
     argp_parse(&argp, argc, args, 0, 0, NULL);
}

the output of the above code looks like:

this is post_doc

  -?, --help                 Give this help list
      --usage                Give a short usage message

this is post_doc

In this case the doc is being printed two times.
As mentioned in the comment by Kenny Stauffer on Bugzilla the first
occurrence of `this is post_doc` is erroneous.

If we change the doc[] in the above code to something else like

static char doc[] = "this is pre_doc\vthis is post_doc";

We get the output like:

this is pre_doc

  -?, --help                 Give this help list
      --usage                Give a short usage message

this is post_doc

[1]: https://sourceware.org/bugzilla/show_bug.cgi?id=19038

Patch

diff --git a/argp/argp-help.c b/argp/argp-help.c
index 3b1727c4aa..ee4d247824 100644
--- a/argp/argp-help.c
+++ b/argp/argp-help.c
@@ -1465,10 +1465,11 @@  argp_doc (const struct argp *argp, const
struct argp_state *state,
   size_t inp_text_limit = 0;
   const char *doc = dgettext (argp->argp_domain, argp->doc);
   const struct argp_child *child = argp->children;
+  char *vt = 0;

   if (doc)
     {
-      char *vt = strchr (doc, '\v');
+      vt = strchr (doc, '\v');
       inp_text = post ? (vt ? vt + 1 : 0) : doc;
       inp_text_limit = (!post && vt) ? (vt - doc) : 0;