Hello Gavin,
Answering to myself: I had been a bit too hasty, patch #7 sent
preivously was working properly, but the naming of variable
_run_tex_in_input_funnies was not corresponding any longer to its
content, so I have slightly modified the code for it to have the same
meaning as before (as in patch #6).
Vincent.
>Hello,
>
>See below...
>
>Le 09/05/2016 21:27, Gavin Smith a écrit :
>> On 9 May 2016 at 20:21, Gavin Smith <gavinsmith0...@gmail.com> wrote:
>>
>>>> I guess that with all those changes, but except for the space that
>>>> is a deep tex engine limitation, only the Devil could find a valid
>>>> filename not working.
>>>>
>>>> Vincent.
>>>>
>>
>> I was going to say, I don't think it's important to support spaces in
>> filenames. (texi2dvi "My Manual.texi" won't work, for example.) The
>> main problem is spaces in higher-level directories, but they
>> shouldn't be a problem any more now we use absolute paths less. I'd
>> say that supporting unusual filenames isn't important. There was
>> already support for tildes (~) in filenames (Debian used them for
>> some reason), but the other characters are easy to avoid by avoiding
>> using them in the filename. Nonetheless it's good that you found a
>> way to support them - it makes texi2dvi slightly more robust for
>> unexpected circumstances. It's good that much of this code isn't used
>> if the filename doesn't contain any unusual characters; this reduces
>> the likelihood of something breaking.
>
>I definitely 200% support your point that we should not change the
>simple \input FILENAME way when there is no need for that. So, I have
>modified the patch so that the condition to trigger the complex way is
>that the filename contain TeX special character, previously that
>condition was that the filename contains non-alphanumeric characters
>which is far broader.
>
>In a nutshell, if the filename contains only alphanumeric or '+' or
>'-', the new patch will just \input FILENAME, but if the filename
>contains '%' or '#', then it will use the TeX tricks.
>
>I also fixed some comments and added '_' into the list of handled TeX
>special characters, for consistency with handling also '$'. Modern tex
>engines seem to already do appropriate catcoding for the math special
>char ($_^) but this behaviour is unspecified, so good to have this
>explicitely.
>
> Vincent.
Index: texi2dvi
===================================================================
--- texi2dvi (revision 7159)
+++ texi2dvi (working copy)
@@ -905,38 +905,86 @@
fi
}
+find_egroup_char()
+{
+ _filename=$1
+ shift
+ while test $# -gt 0; do
+ case "$_filename" in
+ *"${1:0:1}"*)
+ shift;;
+ *)
+ echo "$1"
+ return 0
+ ;;
+ esac
+ done
+ return -1
+}
+
# run_tex - Run TeX as "$tex $in_input", taking care of errors and logs.
run_tex ()
{
make_tex_cmd
- tex_cmd=`(LC_ALL=C
- if (echo "$in_input" | grep [^a-zA-Z0-9/.]) >/dev/null ; then
- # If the filename has funny characters, change the TeX category codes of
- # some characters within a group, and use \expandafter to input the file
- # outside of the group.
+ _run_tex_in_input="$in_input"
+ _run_tex_in_input_funnies=`echo "$in_input" | $SED 's![^} #$%&^_{~]!!g'`
+ if test -n "$_run_tex_in_input_funnies" ; then
+ # If the filename has funny characters, change the TeX category codes of
+ # some characters within a group, and use \expandafter to input the file
+ # outside of the group.
- # Make > an end group character, as it's unlikely to appear in
- # a filename.
- tex_cmd="$tex_cmd '${escape}bgroup${escape}catcode62=2'"
- # make letter { } $ # ~ ^ &
- for w in 123 125 36 35 126 136 26; do
- # TODO: check if letter actually in filename in order
- # to simplifiy TeX command line.
- tex_cmd="$tex_cmd '${escape}catcode$w=11${escape}relax'"
- done
- # Set \toks0 to "\input FILENAME\relax"
- tex_cmd="$tex_cmd'${escape}toks0${escape}bgroup"
- tex_cmd="$tex_cmd${escape}input $in_input${escape}relax>"
- tex_cmd="$tex_cmd${escape}expandafter${escape}egroup"
- tex_cmd="$tex_cmd${escape}the${escape}toks0${escape}relax'"
- else
- # In the case of a simple filename, just pass the filename
- # with no funny tricks.
- tex_cmd="$tex_cmd '${escape}input' '$in_input'"
+
+
+ tex_cmd="$tex_cmd '${escape}bgroup'"
+
+ # seek for _run_tex_egroup_char as a character for egroup (catcode 2)
+ _run_tex_egroup_char=
+ # make other TeX special characters
+ for w in '}125' ' 32' '#35' '$36' %37 '&38' ^94 _95 '{123' '~126'; do
+ case "$_run_tex_in_input" in
+ *"${w:0:1}"*)
+ tex_cmd="$tex_cmd '${escape}catcode${w:1}=12${escape}relax'"
+ ;;
+ *)
+ if test -z "$_run_tex_egroup_char" -a "$w" != ' 32'; then
+ _run_tex_egroup_char="$w"
+ fi
+ ;;
+ esac
+ done
+
+ # If the previous loop has failed to find an egroup character we
+ # try a few more, which don't have any special TeX meaning, but
+ # are unlikely in a file name. This is sure to succeed under MSW
+ # as > < | and " are forbidden in file names.
+ if test -z "$_run_tex_egroup_char"; then
+ _run_tex_egroup_char=`find_egroup_char "$_run_tex_in_input" '>62' '<60'
'|124' '/47' '"34'`
fi
- echo "$tex_cmd")`
+ if test "$_run_tex_egroup_char" != '}125'; then
+ tex_cmd="$tex_cmd
'${escape}catcode${_run_tex_egroup_char:1}=2${escape}relax'"
+ fi
+
+ # Some explicit escaping is needed for the when the FILENAME
+ # contains a ^^XX quadrigram or a simple quote
+ _run_tex_in_input=`echo "$_run_tex_in_input" | \
+$SED 's!\\^\\^\\([0-9A-F]\\{2\\}\\)!\\\\string^\\\\string^\\1!g; \
+s!\\('"'"'\\+\\)!'"'"'"\\1"'"'"'!g'`
+
+ # Set \toks0 to "\input FILENAME\relax"
+ tex_cmd="$tex_cmd '${escape}toks0${escape}bgroup"
+ tex_cmd="$tex_cmd${escape}input' '$_run_tex_in_input'
'${escape}relax${_run_tex_egroup_char:0:1}"
+ tex_cmd="$tex_cmd${escape}expandafter${escape}egroup"
+ tex_cmd="$tex_cmd${escape}the${escape}toks0${escape}relax'"
+ else
+ # Some sh trick is needed in case that the filename contains a simple quote
+ _run_tex_in_input=`echo "$_run_tex_in_input" | $SED
's!\\('"'"'\\+\\)!'"'"'"\\1"'"'"'!g'`
+ # In the case of a simple filename, just pass the filename with no
+ # TeX funny tricks.
+ tex_cmd="$tex_cmd '${escape}input' '$_run_tex_in_input'"
+ fi
+
verbose "$0: Running $tex_cmd ..."
if eval "$tex_cmd" >&5; then
case $out_lang in
